Week beginning 26 October 2026

Commercial implications of this week’s papers

249 of this week’s computer science papers describe a result that could become a product or a service. Open a paper for the summary and the full list of uses.

The headline and the lines under it are written by an AI system from the abstract. They can be incomplete or wrong. The paper is the authority.

Agentic AI predicts crop losses from extreme weather in Florida

From extreme weather to agricultural consequences with agentic AI

Commercial implications: For agricultural insurance providers: Enables more accurate insurance products that price risk and process claims more efficiently for agricultural clients.

Thu 29 OctArtificial Intelligence
The gist
Extreme weather like hurricanes, freezes, and heat can damage crops, but it's hard to predict which crops will be affected and by how much. The authors created WxCrop, a smart system that uses weather data and past agricultural records to estimate crop losses accurately. They tested it in Florida using insurance and production records, showing it can closely predict damages from different weather events. This helps farmers and governments respond faster and better to agricultural disasters.
Open → 2610.40641v1

SonoWorld4D creates immersive 4D audio visual scenes from single videos

SonoWorld4D: From One Video to a 4D Audio-Visual Scene via Diffusion Synchronization

Commercial implications: For virtual reality developers: Enables VR applications and platforms to create immersive content with realistic spatial audio and visuals automatically from video data.

Thu 29 OctComputer Vision and Pattern RecognitionSound
The gist
Generating dynamic scenes that include both visuals and realistic sound is difficult because sound depends on motion and position. The authors introduce a new task to turn a simple video into a 4D scene you can explore, including spatial audio that changes with your viewpoint. Their method recreates a full background, tracks moving sound sources, and synchronizes sounds physically with the scene’s geometry. This lets you see and hear events from new angles in real time.
Open → 2610.40640v1

Computing fair market prices with fees taxes and subsidies included

Finding Competitive Equilibria in Online Marketplaces with Platform Fees, Taxes, and Subsidies

Commercial implications: For online marketplace operators: Enables development of pricing engines that incorporate complex fees and incentives improving marketplace efficiency and revenue.

Thu 29 OctComputer Science and Game Theory
The gist
This paper addresses how online markets involving many buyers and sellers can find fair and balanced prices when fees, taxes, and discounts affect transactions. The authors develop a step-by-step method that adjusts prices in a smart way to quickly find the lowest prices where supply meets demand, even when purchases and sales depend on each other. Their approach also allows marketplaces with complex rules about fees to still find stable deals efficiently. This helps online platforms better manage pricing when extra costs or incentives are involved.
Open → 2610.40626v1

Latent 3d gaussian memory enables fast spatially consistent infinite worlds

L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian Memory

Commercial implications: For game developers: Enables faster and more resource-efficient real-time game worlds that retain consistency, enhancing user experience with scalable content.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Generating virtual worlds that stay consistent as you explore them is hard, especially if you want the world to update in real time. The authors introduce a new way to remember and represent 3D scenes using special blobs called latent 3D Gaussians that grow with the world size, not how long you explore. Their method lets the system create detailed, consistent imagery quickly without needing to reprocess images repeatedly, making it fast enough for real-time use. Experiments show their approach is much faster and uses fewer resources than earlier methods.
Open → 2610.40614v1

Game model separates combat rules from video for better gameplay

Fight World Model: Customizable Combat through Visual Interaction Proxies

Commercial implications: For video game developers: Enables selling more flexible and modifiable fighting game engines supporting new mechanics and character designs.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Video game characters fighting on screen need a system that keeps track of health and hit reactions properly. The authors introduce Fight World Model, which divides the game into two parts: one that strictly follows the rules and keeps track of fighting details, and another that creates the visuals. This lets players see consistent hit effects and health changes, and allows new fighting abilities without retraining the visual system. Their method works well with many characters and arenas and beats earlier models at keeping things accurate.
Open → 2610.40607v1

Quality checker improves trust in dental X-ray segmentation results

Beyond Ground Truth: A Heuristic Unsupervised Quality Assurance Engine for Diagnosing Calibrational Blindness in Dental Segmentation

Commercial implications: For medical device developers: Enables product features that alert users to unreliable dental segmentations, enhancing clinical confidence and safety.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Dental X-ray images need to be segmented accurately to help dentists make diagnoses, but deep learning models often give wrong segmentations with high confidence, which is misleading. The authors created a tool called HUS-QA that checks the quality of the segmentation without needing a correct answer to compare against. This tool looks at features like confidence, edge sharpness, and shape connectivity to guess how good the segmentation is. They found that this method correlates well with actual segmentation quality and works better than just trusting the model's confidence score. This can help doctors trust automated diagnoses more in real clinical settings.
Open → 2610.40587v1

Multi teacher approach improves training of large language models

Every Token Needs Its Own Teacher: Multi-teacher Adaptive On-policy Self-distillation

Commercial implications: For ai service providers: Enables production-quality AI services with more accurate model outputs, improving competitive advantage in AI-assisted reasoning tools.

Thu 29 OctComputation and Language
The gist
Teaching a large language model is tricky because different parts of its responses might need different guidance. The authors found that instead of using one teacher for an entire response, it’s better if each word in the response can choose its own teacher. They created a new method where multiple teachers help the model, selecting the best one for each word during training. This approach made the models better at solving math problems in tests.
Open → 2610.40574v1

Farms builds a trustworthy knowledge base for automated computing science

FARMS: A Verifiable Foundation for Automated Computational Science

Commercial implications: For automated ai tooling teams: Enables commercial AI platforms to offer provably correct computational method selection for scientific software, increasing trust and efficacy.

Thu 29 OctArtificial IntelligenceLogic in Computer Science
The gist
Scientists want to use AI to help pick and improve math methods for computers to solve scientific problems faster. The authors created a tool called FARMS that turns complex math statements into a form AI can understand and trust, linking these statements back to their original sources and proofs. FARMS was tested on many math problems related to linear algebra and differential equations, and it successfully verified most of them efficiently and cheaply. This helps make automated problem-solving in science more reliable and easier to reuse. The authors show examples where their system connects math knowledge with real computational tasks like heat flow and wave response.
Open → 2610.40560v1

LucidLyra makes 3D scenes from videos directly searchable with objects

LucidLyra: Feed-Forward Generative 3D Scene Reconstruction and Joint Understanding via Multi-Task Distillation

Commercial implications: For game developers: Enables game studios to generate detailed, interactive 3D scenes rapidly from video footage, reducing asset creation costs.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Making detailed 3D scenes from videos usually needs lots of images or slow processing steps. The authors created LucidLyra, a system that builds 3D scenes with object details and categories directly from video data, without extra manual work or complicated setups. It uses knowledge from multiple AI models to understand scene shapes, colors, and objects together, allowing users to look around and find things in 3D worlds quickly. This can help robots or virtual tools work in these scenes more effectively.
Open → 2610.40495v1

Agentic University supports personalized learning with collaborative agents

Agentic University: Collaborative Agents for Personalized Learning

Commercial implications: For e-learning platform developers: Enables creation of personalized tutoring software with enhanced learner engagement and improved assessment outcomes, attractive for education technology products.

Thu 29 OctArtificial IntelligenceComputation and Language
The gist
Learning can be more effective when teaching adapts to what each person needs, but making this work smoothly is hard. The authors present Agentic University, a system where learners team up with software agents that help plan learning goals, activities, and tests, and adjust this plan based on how the learner is doing. This setup helps learners improve more and feel better supported than traditional methods. The system keeps the learner involved in deciding the learning path and uses feedback to guide what comes next.
Open → 2610.40492v1

FIRME improves linking investor knowledge across different sources and times

FIRME: Registry-Anchored Memory Graphs for Evolving, Heterogeneous Investor Knowledge

Commercial implications: For business intelligence developers: Enables sale of advanced analytics platforms that integrate multiple investor data types with higher accuracy and provenance tracking.

Thu 29 OctArtificial IntelligenceComputation and Language
The gist
Financial AI systems that track investor information often struggle to keep track of who or what is being talked about when data uses different names or descriptions. The authors describe FIRME, a system that merges various pieces of investor data into consistent, linked records while keeping track of where and when the data came from. This helps create a clearer picture by connecting conversations and financial signals without losing track of their source. Tests show FIRME matches information more accurately and maintains better evidence chains than previous methods.
Open → 2610.40482v1

Spatial audio improves tracking of sound-producing people in videos

AVPS-360: A Spatial Audio-Visual Benchmark for Instance Segmentation and Tracking of Sound-Producing Persons

Commercial implications: For augmented reality developers: Enables AR products to accurately identify and track sound sources in 3D space, enhancing interactive experiences.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Tracking who is making sounds in videos is hard when only regular audio from one microphone is used. The authors created a large new collection of videos that use special 360-degree sound recording to better locate where sounds come from. They also made a new method that uses this spatial sound together with video images to better identify and follow people making sounds like talking or clapping. Their tests show that older methods don’t work well on these long videos with breaks in sound, but the new method does better at both finding and linking the correct people to their sounds.
Open → 2610.40480v1

Simulation training improves language models for psychiatric interviews

Training language models in simulation for psychiatric intake

Commercial implications: For mental health technology developers: Enables development of AI tools that assist clinicians with structured psychiatric interviews, improving care delivery and workflow.

Thu 29 OctArtificial IntelligenceComputation and Language
The gist
Psychiatric intake interviews are important and complex conversations between doctors and patients to understand mental health. The authors trained a language model by having it practice simulated interviews using reinforcement learning, which helped it get better at asking questions, making diagnoses, and planning care. When tested in role-played scenarios with real psychiatrists, the trained model performed better than the original. The researchers also created tests to check if the model safely recognizes urgent patient risks and found the training reduced missed dangers.
Open → 2610.40476v1

Physically plausible object insertion improves lighting and perspective matching

PhysInsert: Physically Plausible Object Insertion with Geometric Priors and Bidirectional Object-Scene Interactions

Commercial implications: For augmented reality developers: Enables AR apps to place virtual objects realistically in live environments, improving user experience and engagement.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Inserting objects realistically into images is hard because objects need to match lighting, position, and how they interact with the surrounding scene. The authors created a new dataset using 3D software to capture exact changes from different viewpoints and lighting, helping train models better. They built a system called PhysInsert that uses geometry and image-text guidance to place objects in images more naturally, adjusting their pose and lighting. Tests and user studies showed their method makes inserted objects look more believable and better integrated with the scene.
Open → 2610.40453v1

Confidence alignment improves reinforcement learning in language models

SCARL: Sequence Confidence-Aligned Reinforcement Learning

Commercial implications: For ai product developers: Enables more reliable AI-generated content for product features requiring high accuracy and user trust.

Thu 29 OctMachine LearningArtificial Intelligence
The gist
Language models learn by getting rewards when their answers pass a test, but sometimes they are confident about wrong answers or uncertain about right ones. The authors noticed this mismatch between how confident the model feels and how good its answer actually is. They created a new method called SCARL that adjusts the learning process so the model's confidence better matches the correctness of its answers. This helps the model explore more and improves its performance on reasoning tasks.
Open → 2610.40407v1

First-order method improves design of non-truthful auctions with multiple bidders

First-Order Bilevel Optimization for Non-truthful Auction Design

Commercial implications: For online advertising platforms: This enables building better auction-based ad platforms that extract higher revenues by exploiting bidder strategies at scale.

Thu 29 OctComputer Science and Game Theory
The gist
Designing auctions where bidders do not just tell the truth about their values is hard because bidders change their bid strategies as the auction changes. The authors treat this as a game between an auction designer and many strategic bidders and create a new method that uses only gradient information to optimize auction revenue. Their approach can find good auction designs faster than older methods and works well even when there are many bidders or multiple items for sale. It also finds auctions that earn almost as much as those with truthful bidding, offering a new way to automate the design of complex auctions.
Open → 2610.40403v1

Structured textbook data improves math proof learning in language models

Scaling Mathematical Proof Supervision from Structured Textbooks

Commercial implications: For educational software developers: Enables development of advanced math tutoring AI that adapts to curriculum structure and reasoning steps, improving learner outcomes.

Thu 29 OctComputation and LanguageArtificial Intelligence
The gist
Mathematical problems and their solutions are often taught in an organized way within textbooks, but machine learning models usually see these problems and solutions as separate, unrelated examples. The authors created a large collection called TOME-Math that keeps all the organization and connections found in textbooks, like chapters and how different propositions relate to each other. By using this well-structured data to train language models, the authors boosted the models' ability to reason mathematically by a significant margin. This work shows that preserving the structure in learning materials helps models understand math better.
Open → 2610.40394v1

Agentic llm predicts pediatric diabetes outcomes better than classic machine learning

Agentic LLM reasoning in pediatric diabetes outcome prediction from multimodal medical records

Commercial implications: For developers of clinical decision support: Allows creation of AI-based medical software that enhances doctor decision-making with transparent, reliable clinical predictions.

Thu 29 OctComputation and Language
The gist
Predicting health outcomes for children with diabetes is hard because there are fewer patients and uncommon events. The authors used a special computer program called CEDAR that reads different types of medical information and explains its reasoning. This program predicted future health problems more accurately than traditional computer methods in young people with type 1 and type 2 diabetes. Doctors who looked only at the program’s explanations agreed with its predictions in most cases.
Open → 2610.40379v1

Cognitive resource management helps AI agents handle long tasks better

Don't Lose the Thread: Empowering Long-Horizon LLM Agents with Cognitive Resource Self-Allocation

Commercial implications: For ai product developers: Enables building advanced AI software that solves longer, more complex tasks accurately while reducing compute costs.

Thu 29 OctComputation and Language
The gist
Large language model agents can get confused when working on long tasks because their memory gets crowded with too much information. The authors introduced a technique called CORAL that helps these agents organize what they remember by saving important facts and clearing out cluttered info. This lets the agents start fresh when needed but still remember key details. Their method improved performance on tough long-step tasks while using much less computing effort.
Open → 2610.40350v1

Agentic generation creates 400K articulated assets for simulation

RFW-A: Agentic Procedural Generation of 400K Articulated Assets over 500 Categories

Commercial implications: For simulation platform providers: Supports building and selling advanced simulation content and training environments for robotics customers.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Creating detailed 3D objects that can move and be used in simulations usually takes a lot of human effort. The authors found a way to teach AI agents to build general instructions (programs) for making whole categories of these objects. Once made, these instructions can quickly create many different objects without extra AI calls, saving time and cost. Using this method, they generated over 400,000 simulated articulated objects in 500 categories, which can help improve robots' understanding and interaction with real-world objects.
Open → 2610.40333v1

Guideline decision graph boosts radiology case accuracy and speed

Radiology Guideline-derived Decision Graph for Efficient Clinical Decision Support

Commercial implications: For clinical decision support product teams: Enables building and selling diagnostic support software that more reliably follows official radiology guidelines, appealing to hospitals seeking better compliance.

Thu 29 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Medical guidelines for interpreting imaging tests can be complicated and scattered across many sections and tables. The authors created a system called GDGraph that breaks down these guidelines into clear decision steps. This system helps computers follow the guidelines when making recommendations, making their answers more accurate and easier to understand. Tests showed GDGraph outperforms other AI methods in interpreting kidney, ovary, and thyroid imaging cases quickly and correctly.
Open → 2610.40327v1

Latent path exploration improves reasoning in language models

Latent Path Exploration via Potential-Guided Subtrajectory Balance

Commercial implications: For automated tutoring platforms: This paper enables smarter AI tutors that reason more effectively, potentially increasing their market value and adoption.

Thu 29 OctComputation and LanguageMachine Learning
The gist
Language models can solve problems by thinking through steps inside their 'mind' without saying each step out loud. The authors found a way to better explore different possible thinking paths using a technique that scores each step based on how well it might lead to the right answer. Their approach helps the model try many paths and pick the better ones, especially in math and logic puzzles. This leads to more correct answers, even when tested on harder or unfamiliar problems.
Open → 2610.40283v1

Coding agents improve long task handling with complex rules

When Requirements Interact: Scaling Long-Horizon Coding via Interdependent Specifications

Commercial implications: For automated software tool vendors: This paper enables vendors to build smarter AI coding tools that meet complex real-world developer requirements, improving product marketability.

Thu 29 OctSoftware EngineeringComputation and Language
The gist
Software development often requires meeting many important rules at once, like making sure parts fit together and errors are handled properly. Current methods usually only focus on one main goal and miss these tricky rules. The authors created a system called CodeSpecWeave that helps generate tasks combining many connected rules and checks solutions carefully. This system helped improve coding agents' abilities on a big set of tasks and made a new, tough benchmark to test these skills. Their work shows better ways to train software tools to handle complicated programming needs.
Open → 2610.40256v1

Image watermarking method resists attacks on preexisting pictures

Attack-Resilient Multi-Bit Image Watermarking Framework via Latent Coordinate Sign Clamping

Commercial implications: For video game asset managers: Enables selling watermark solutions that secure game assets against piracy, leveraging this paper’s attack-resistant watermarking framework.

Thu 29 OctCryptography and Security
The gist
Protecting image ownership is hard because methods often can't handle existing images, store many bits of data, or resist deliberate tampering. The authors created a watermarking method that can hide multi-bit messages in any image without needing retraining, and still keep the image looking natural. Their method also includes advanced ways to decode the message even if noise or attacks try to remove it. It works well on natural and AI-generated images and can detect attempts to forge or alter the watermarks.
Open → 2610.40204v1

Advertising images created using skill-guided multi-agent approach

AdMuse: Skill-Guided Creative Advertising Image Generation

Commercial implications: For digital marketing platforms: Enables marketing platforms to offer AI-powered customized ad image generation that improves campaign effectiveness.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Creating good advertising images is hard because they need to catch attention and clearly show what a product is about. The authors propose AdMuse, which breaks down advertising ideas into reusable 'skills' that guide image creation across different products. Instead of copying examples, AdMuse uses many skills to suggest ideas, picks the best ones, and refines images using feedback from vision and language models. This makes ad creation a step-by-step process rather than just writing one prompt. Tests showed AdMuse makes attractive ads that match what products stand for.
Open → 2610.40190v1

Multi-channel sensors improve camera color correction beyond color temperature

Beyond CCT: Multi-channel SPD recovery for per-shot camera color correction

Commercial implications: For smartphone manufacturers: Enables selling smartphones with superior camera color fidelity in varied lighting environments, differentiating from competitors.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Modern cameras usually adjust colors based on how warm or cool a light looks, but this method can mix up different lights that look similar. The authors show that by measuring more details about the light’s spectrum using extra sensors, cameras can better figure out the exact color of the light. This means color corrections are more accurate, especially for tricky light sources like LEDs and mixed lighting. Their method works well on both synthetic and real examples, even with some sensor noise.
Open → 2610.40175v1

Event cameras reconstruct color from monochrome brightness changes

EventSeeColor: Learning to Reconstruct Color from Monochrome Events

Commercial implications: For mobile camera developers: Allows new color event camera products or software for consumer smartphones that work better in difficult lighting.

Thu 29 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Traditional event cameras only capture changes in brightness, so they produce black-and-white images. The authors show a method to guess what the colors are from these black-and-white event signals. Their method uses both the grayscale images and the original event data together to colorize scenes more accurately. This helps reveal differences in appearance that black-and-white images miss, useful for seeing details in tricky lighting.
Open → 2610.40166v1

Framework helps interpret user intents and find cloud services despite conflicts

A Negotiation-Aware Hybrid LLM Framework for Intent Interpretation and Service Discovery

Commercial implications: For automated service brokers: This paper enables products that interact with users in natural language to refine service requests and select cloud offerings automatically.

Thu 29 OctArtificial IntelligenceSoftware Engineering
The gist
People often describe what they want from cloud services in natural language, which can be unclear or contradictory. The authors propose a system that can understand these intent statements, identify any conflicts, and then find appropriate cloud services that match the user’s needs. If conflicts or unsatisfiable requests arise, the system negotiates with the user in natural language to clarify and resolve them. This approach improves how well the system detects conflicts and finds suitable services compared to simpler existing methods.
Open → 2610.40161v1

Volumetric model improves abdominal tumor segmentation using reports

RepConTUR: A Report-Conditioned Volumetric Multimodal Model for Abdominal Tumor Segmentation

Commercial implications: For radiology software developers: This enables commercial imaging software to provide more precise tumor outlines for radiotherapy by jointly analyzing scans and diagnostic reports.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Correctly outlining tumors in scans is important for planning treatments but automatic tools often make mistakes by marking wrong areas. The authors developed a new 3D computer model called RepConTUR that uses the text written by doctors about the scan to better find tumors in organs like the liver, kidney, and pancreas. Their approach combines the detailed CT scan images with the narrative reports to segment tumors more accurately compared to existing methods. They also created new tasks and data that help train and measure this combined image-and-text understanding.
Open → 2610.40094v1

Language model adapts better by routing domain experts dynamically

Upcycling Language Models for Multiple Specialized Domains

Commercial implications: For product managers in ai services: Supports commercial AI products that handle diverse domains with higher accuracy and efficiency.

Thu 29 OctComputation and LanguageArtificial Intelligence
The gist
Large language models sometimes struggle when asked to perform tasks from many specialized areas at once. The researchers propose a way to add parts, called experts, that each focus on a specific domain and a router that learns to pick the right expert for each input. This method lets the model perform well across different domains without confusion. They show that just a small number of examples can teach the router to choose experts accurately, improving overall results while keeping the original abilities of the model.
Open → 2610.40091v1

Agent process evidence improves safety and accountability in ai software development

Agentic process evidence

Commercial implications: For financial compliance officers: Enables regulatory-compliant AI governance tools for industries requiring strict audit trails.

Thu 29 OctArtificial IntelligenceSoftware Engineering
The gist
AI agents are now helping write and manage software but their work disappears after each session, making it hard to track what they did or hold them responsible. This causes problems for security and compliance because we can't see their decisions or ensure they follow company rules. The authors propose Agentic Process Evidence (APE), a system that securely records what AI agents do and links their actions to outcomes. This helps companies keep control, audit, and govern AI contributions like they do for human developers.
Open → 2610.40071v1

Improving 3D CT images from cone beam scans using physics constraints

EPC-RawDC: Equivariant Physics Constrained 3D Diffusion with Raw Data Consistency for CBCT to CT Image Synthesis

Commercial implications: For radiation therapy device manufacturers: Enables devices to sell advanced CBCT scanners with built-in CT quality enhancement features appealing to cancer treatment centers.

Thu 29 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Cone beam CT scans are often used to guide cancer treatments but can have blurry or inaccurate images. The authors created a new AI method that helps turn these scans into clearer, more accurate CT images by using knowledge about how the scanner works. Their approach combines 3D image learning with physics rules about the scanning process, making the results better than previous methods. This method was tested on real and simulated scans from different machines and showed more precise and reliable images.
Open → 2610.40046v1

UPPAAL gains interactive help from large language models for system modeling

UPPAAi: Towards LLM-Assisted Modeling, Verification, Analysis and Synthesis

Commercial implications: For software development toolmakers: Enables new AI-powered formal verification products that improve user experience and broaden customer base.

Thu 29 OctSoftware EngineeringArtificial IntelligenceLogic in Computer Science
The gist
Modeling complex systems like timed or hybrid automata can be hard and often requires expert knowledge. The authors created UPPAAI, a tool that integrates a large language model (LLM) with the existing UPPAAL software to help users build, explain, and check these models through a conversation. The LLM suggests and explains steps while UPPAAL verifies correctness, making model building more approachable and interactive. This new tool was tested on different case studies, showing that LLMs can proactively offer helpful features even when users don’t ask for them.
Open → 2610.40042v1

Optimized urban vegetation layouts balance shade and aesthetics

Greenscapes: Optimized Vegetation Configurations in Cities

Commercial implications: For game developers: Enables advanced city environment generation tools that enhance realism and environmental effects for games.

Thu 29 OctGraphicsComputer Vision and Pattern Recognition
The gist
Vegetation in cities helps with shade, sound control, climate, and makes places look nice. The authors created a new way to decide where to plant trees and shrubs in 3D city views. They use a tool that understands pictures and language to suggest plant positions, then improve these placements with a computer method aiming for better shade and noise control. This method also considers how the scene looks from different views to keep it attractive. It helps design city green spaces that are both useful and good to see.
Open → 2610.40007v1

Hyperparameter tuning improves atrial fibrillation detection accuracy

Enhancing Atrial Fibrillation Detection Through Hyperparameter Optimization

Commercial implications: For mobile health app developers: Enables creation of consumer health monitoring apps that provide more accurate AF detection, increasing market appeal and health impact.

Thu 29 OctMachine Learning
The gist
Atrial fibrillation is a heart condition that can be detected using computer models analyzing heart rate data. The authors studied how adjusting settings called hyperparameters can improve the accuracy of different computer models. They tested several machine learning and deep learning methods and found that one method, using Hyperband to tune an XGBoost model, worked best. This helps build better tools for detecting atrial fibrillation in real-world medical data.
Open → 2610.40001v1

New method improves recommendation queries using retrieval feedback

R2QRec: Learning Recommendation Queries from Retrieval Feedback

Commercial implications: For e-commerce developers: Improves commercial recommendation engines by refining query generation, leading to better item discovery and increased sales.

Thu 29 OctInformation Retrieval
The gist
Recommendation systems suggest items you might like, but it's tricky to turn past user choices into search queries that find those items again. The authors present R2QRec, a method that learns to create good search queries without needing labeled examples of queries. It uses a fixed search system to check which queries find the right items, then improves queries based on how well they rank those items. This approach helps the system get better recommendations across different datasets without changing the search engine itself.
Open → 2610.39984v1

Text-to-image models generate more varied images without losing quality

TokenRepel: Generating Diverse Image Sets Without Compromising Quality

Commercial implications: For advertising content teams: Enables creating varied advertising campaigns efficiently using text-to-image tools, improving customization and speed.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Sometimes, when computers make pictures from words, they create many images that all look very similar. The authors found a way to help these models produce a wider variety of images for the same text prompt while keeping the pictures looking good. They do this by making the computer representations inside the model push away from each other so the images differ more. They also make sure that the changes don’t make the pictures worse by keeping the computer’s predictions in a high-quality zone. This method works well with different types of text-to-image models.
Open → 2610.39913v1

Ink guides smart assistants for organizing ideas in digital workspaces

Inkitiate: Initiating and Steering Agent Assistance through Ink in a Spatial Workspace

Commercial implications: For productivity software developers: Enables creation of advanced AI-augmented note-taking or workspace software that improves user interaction through ink input.

Thu 29 OctHuman-Computer Interaction
The gist
Creating and organizing ideas often involves handwriting and drawing, but people sometimes need help managing their materials. The authors studied how using handwriting (ink) can start and control a helpful assistant in a digital workspace. They built a system called Inkitiate, where users can use ink to bring in information, think with it, and organize their space while guiding the assistant’s actions. Tests showed users found it usable and easy to work with, combining writing and drawing to ask for help effectively.
Open → 2610.39905v1

Orientation maps speed robotic needle entry point planning

Compressed Orientation Reachability Maps for Sub-Second Robotic Needle Entry Point Planning

Commercial implications: For medical device teams: This paper enables real-time needle trajectory planning software valuable to companies making robotic surgery devices.

Thu 29 OctRoboticsComputational Geometry
The gist
Planning where a robot can safely insert a needle into the liver during surgery takes a lot of time, slowing down the process. The authors created a special map that shows which needle angles and positions work for the robot without having to recalculate each time. This map is stored in a small, easy-to-access file, letting the robot find a good entry point very quickly. Their tests showed it speeds up planning from over 13 seconds to less than half a second, helping surgeries happen faster and more smoothly.
Open → 2610.39879v1

Augmented video models keep scenes and movements consistent longer

GROME: GROunding Factorized MEmory into Video World Models

Commercial implications: For video game developers: Enables game companies to build more immersive and consistent dynamic environments by extending video memory efficiently.

Thu 29 OctComputer Vision and Pattern RecognitionMachine Learning
The gist
Keeping videos consistent over long periods is hard when scenes and moving objects change outside the usual short memory window. The authors introduce GROME, a method that separates static parts like backgrounds from moving objects in memory. This helps the video model remember scenes and movements better over time without retraining. It also allows for more flexible video extensions by plugging in new observations without extra training.
Open → 2610.39812v1

Small language models guided to teach grammar in language learning

Grammar Control for Language Learning with Sequential Monte Carlo and Small Language Models

Commercial implications: For mobile app developers: Enables on-device language learning apps with strong grammar control without needing large cloud models, improving speed and privacy.

Thu 29 OctComputation and Language
The gist
Learning a new language is easier when learners get practice with just the right grammar. The authors found a way to guide small, simple language models to produce grammar learners need, using a technique called sequential Monte Carlo. This approach lets these smaller models perform as well or better than much bigger models without needing tons of extra training data. They also showed how to create new grammar guides automatically, making it work for many grammar points without extra effort.
Open → 2610.39800v1

Hybrid memory boosts long-term dialogue agent accuracy

Long-Term Agent Memory through Complementary Curation and Consolidation

Commercial implications: For chatbot developers: Enables more reliable long-term memory in commercial chatbots, improving customer interactions and retention.

Thu 29 OctArtificial IntelligenceMultiagent Systems
The gist
Memory systems help AI assistants remember past conversations and facts to respond better over long chats. The authors designed a memory setup that mixes two types: one that stores specific events and another that keeps organized facts and relationships. They also created a way to keep this memory efficient over time. Their tests show this design works better than existing ones on two standard challenges.
Open → 2610.39790v1

Streaming system protects voice privacy and removes personal info fast

Toward Complete Speaker Anonymization: a Streaming Cascaded System for Voice Privacy and PII Removal

Commercial implications: For voice assistant developers: Enables selling privacy-focused voice assistant software that anonymizes user speech and removes private details on the fly.

Thu 29 OctSound
The gist
This paper shows a new system that can protect your voice identity when speaking, by removing personal information in real time. The system combines speech recognition, language understanding, and voice synthesis in a sequence to keep your spoken content clear while hiding who you are and sensitive details. The authors demonstrate it works almost as well as slower offline systems but with much less delay. They also provide ways to control how much content or personal info gets protected.
Open → 2610.39754v1

Uniffordinteractor enables more realistic human and object interactions in 3d scenes

UniffordInteractor: Reasoning the Unified Affordance for Human and Multi-Object Interaction

Commercial implications: For virtual reality developers: Enables VR applications and games to simulate complex human-object interactions improving user immersion and experience.

Thu 29 OctGraphics
The gist
When people move and interact with many objects at once, like picking up items or using tools together, computers find it hard to predict these actions realistically. The authors created a new method called UniffordInteractor that first imagines how objects can be used together and then simulates how a person would interact with them over time. This approach looks at both how humans touch objects and how objects relate to each other, improving the realism of these interactions. Their tests show it works better than previous tools at modeling complicated scenes with many objects.
Open → 2610.39745v1

Speech genre affects training of turn-ending detectors in conversations

The Importance of Speech Genre for End-of-Turn Detector Training

Commercial implications: For voice assistant developers: Enhances speech interaction quality for commercial voice-activated products by enabling smoother turn-taking.

Thu 29 OctArtificial Intelligence
The gist
Detecting when someone finishes speaking is important for smooth conversations, especially between humans and machines. This paper shows that training these detectors on real human conversations works better than training on one-turn exchanges with machines. The researchers found that models trained on conversations make fewer mistakes, especially with pauses and hesitations. This advantage only appears when testing on conversational speech, meaning most past work missed this benefit by focusing only on single-turn data.
Open → 2610.39742v1

Confidence filtering affects accuracy of video-based gait assessments

Confidence Filtering in Video-Based Gait Assessment: Measurement Error versus Selection Shift

Commercial implications: For medical device developers: This paper guides production of more accurate gait analysis devices for clinical customers by revealing filtering-induced biases.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Video systems that analyze walking patterns often ignore low-confidence data points to reduce measurement mistakes. The authors compared this approach to random data selection in people with Parkinson’s and another condition. They found that filtering by confidence can slightly change the movement results, mixing true errors with selection bias. Filtering helps only if confidence relates mostly to error, but if confidence is linked to the movement itself, it can shift the results and introduce bias. In their Parkinson’s data, confidence had little relation to movement, so filtering was mostly harmless if unbiased windows were kept.
Open → 2610.39704v1

Smartphone indoor tracking improves by using layered sensor context

Smartphone Indoor Localization Enhanced by Hierarchical Multi-Sensor Context Recognition

Commercial implications: For mobile app developers: This paper enables selling reliable indoor navigation features to app users who need accurate position tracking inside buildings.

Thu 29 OctMachine LearningRobotics
The gist
Indoor phone location is tricky because walls and floor changes confuse normal tracking. The authors show a way to use multiple phone sensors together that first figure out where you are and what you’re doing. Then their method corrects tracking errors based on these layered clues, like whether you’re walking in a corridor or carrying your phone differently. Tests showed this approach helps smartphones know their position more accurately inside buildings.
Open → 2610.39691v1

Deep neural networks improve atrial fibrillation detection accuracy

A New AFIB Detection Method with Deep Neural Networks

Commercial implications: For medical device engineers: Enables more accurate AFIB detection in wearable and clinical monitors, creating better diagnostic products for healthcare providers.

Thu 29 OctMachine Learning
The gist
Atrial fibrillation is a heart condition where the heartbeat becomes irregular, which can be dangerous if not detected early. The authors developed a new way to find this irregular heartbeat by using deep neural networks that learn from a special type of heart rate data, making detection more precise. They tested their method on well-known heart signal datasets and found it works better than previous methods. This could help doctors spot atrial fibrillation more reliably.
Open → 2610.39665v1

Data stories shaped in real time by audience responses

Towards Audience-Driven Real-Time Data Storytelling

Commercial implications: For digital marketing teams: Allows selling interactive marketing tools that customize content flows for targeted audience segments using live data.

Thu 29 OctHuman-Computer Interaction
The gist
Sometimes when people share data stories, the audience just watches quietly. This paper shows how the story can change based on what the audience answers in real time. The authors created a system where the presenter prepares a branching story with questions, and the audience’s answers guide which parts of the story get told next. This makes the story fit what the audience cares about and understands.
Open → 2610.39658v1

Colonoscopy polyp tracking improves without manual bounding box labels

Annotation-Free Representation Learning for Polyp Retrieval in Colonoscopy Videos

Commercial implications: For medical imaging software developers: Enables development of improved colonoscopy image tracking software sold to healthcare providers, reducing annotation costs.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Colonoscopy videos show polyps from different angles, helping doctors understand them better. But to find and track the same polyp in many frames, computers usually need human-drawn boxes around polyps, which is slow and costly. The authors developed a method that learns to find and follow polyps over time using raw videos alone, with no need for these labels. This approach works better than existing methods and can find more images of the same polyp even from just a few examples.
Open → 2610.39648v1

Unified virtual try-on system works for many cloth and person scenarios

FlexTry-On: A Time-Efficient Framework for Multi-Scenario Virtual Try-on

Commercial implications: For clothing retailers: Enables a multi-scenario virtual try-on product appealing to diverse customers with reduced technical complexity.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Virtual try-on technology lets people see how clothes look on them without physically wearing them. Most systems only work for one type of situation, like putting clothes on a single person or swapping outfits between two people. The authors created FlexTry-On, a single system that can handle many different types of virtual try-on tasks quickly and without needing separate setups. This system uses a smart method to understand clothes and people together and works well on tests for different cases. It can even handle new situations it hasn't seen before without extra changes.
Open → 2610.39646v1

Open problems in research fields tracked precisely with a layered knowledge graph

Rebuilding a Field's Open-Problem Catalog from the Literature: A Two-Layer Knowledge Graph, and Why Status Must Stay Proposed

Commercial implications: For scientific publishing platforms: Enables building advanced research discovery tools that deliver curated problem updates and evidence to users.

Thu 29 OctDatabasesArtificial IntelligenceDigital Libraries
The gist
Many research areas list their unsolved problems only in text documents, which makes it hard to know who is working on what or if problems have been solved recently. The authors built a system that turns these open problems into data by combining a hand-curated list with evidence from scientific papers in one searchable graph. They used an AI model to extract problem details and status from papers but found that automatic claims about problem solutions are often incorrect and must be checked by humans. Their work shows it’s important to keep humans in the loop when updating problem statuses, and they discuss challenges like matching entities and verifying claims.
Open → 2610.39641v1

Conversational and dashboard explainability tools show mixed benefits in healthcare AI

Evaluating Conversational Explainability in Healthcare: Comparing the Explainability Assistant and a Traditional XAI Dashboard

Commercial implications: For healthcare software developers: Enables development of commercial AI explanation products tailored for healthcare providers to improve trust and understanding.

Thu 29 OctHuman-Computer Interaction
The gist
Explaining how AI systems make decisions can help healthcare workers trust and use them better. The authors compared a chatbot-style AI explainer with a traditional dashboard that shows detailed visual explanations, both using the same heart disease prediction model. They found that each interface had its own strengths: the chatbot was easier to start using and gave patient-specific answers, while the dashboard better showed overall and detailed information at once. Users suggested combining both styles to get the best of each in real clinical settings.
Open → 2610.39634v1

Interactive video worlds improve accuracy by predicting the next scene

FutureWorld: A Plug-and-Play Causal Next-State Predictor for Interactive Video World Exploration

Commercial implications: For game developers: Enables game engines to produce more stable and consistent interactive video worlds, improving player experience and reducing graphical errors.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Interactive video systems try to create scenes that change naturally when users take actions, but mistakes can build up over time, making the video blurry or inconsistent. The authors found that relying on past generated images to predict future scenes causes errors to accumulate. To solve this, they propose a method that predicts the next scene before generating it, which helps guide the video creation and reduce error growth. Their method works with different video generators and improves consistency in long interactive videos.
Open → 2610.39619v1

Dominance guided method improves medical severity grading from sparse comparisons

DOSE: Dominance-Guided Ordinal Severity Estimation from Sparse Relative Annotations

Commercial implications: For medical ai developers: Enables development of diagnostic tools with improved severity estimation for sale to healthcare providers.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Medical images are often rated using fixed categories like mild or severe, but these categories have a natural order that many methods ignore. The authors propose a method called DOSE that learns how severe a medical condition is by comparing pairs of images instead of needing exact labels. DOSE creates a global ranking from these pairwise comparisons and uses this to assign scores, which it then converts to the usual categories without needing labeled examples. This approach worked better than previous methods on multiple medical image datasets.
Open → 2610.39584v1

Faster text to image generation with one step adversarial training

ROAD: Modern Adversarial Training for Simple One-Step Distillation at Scale

Commercial implications: For ai service providers: This paper enables faster image generation products for clients needing quick and cost-effective AI art solutions.

Thu 29 OctComputer Vision and Pattern RecognitionMachine Learning
The gist
Generating images from text usually takes a lot of time and computing power. The authors introduce ROAD, a simpler way to speed up this process by training in just one step using a special technique called adversarial distillation. They improve the model with new components that better understand both images and text, and adjust training details to avoid common problems. This method achieves high-quality image results faster and with less computing effort than previous approaches.
Open → 2610.39539v1

New method completes missing brain tumor MRI sequences preserving 3D details

MOSAIC-Flow: Orthogonal-Axis and Status-Conditioned Fusion for Missing-Sequence Completion in Brain Tumor MRI

Commercial implications: For medical imaging software developers: Enables software products offering automated MRI sequence completion, improving clinical workflow and imaging consistency.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
MRI scans of brain tumors often have missing image types, which makes it hard to fully understand the tumor. The authors developed a new method called MOSAIC-Flow that can fill in these missing images while keeping fine 3D details intact. Their approach processes each MRI type separately, looks along three different body directions, and adapts based on which images are available. Tested on a dataset of brain tumor MRIs, MOSAIC-Flow showed consistent quality in completing different missing image types, especially structural MRI scans.
Open → 2610.39534v1

Code model hidden states reveal and enable program edits using static analysis

Program Bindings Inside Code Models

Commercial implications: For integrated development environment developers: This paper enables creation of smarter coding assistants that can dynamically edit code logic, appealing to IDE vendors seeking advanced AI features.

Thu 29 OctSoftware Engineering
The gist
Code language models can understand and fix programs, but it’s unclear how they represent a program’s structure inside their memory. The authors checked if information from standard program analysis tools, like how a program’s pieces relate, can describe and even change this hidden memory without retraining the model. They found that using this program structure helps keep the model’s answers correct better than simple position-based descriptions in most tested code models. They also showed they can edit parts of a program inside the model’s memory to make it behave as if the program itself changed.
Open → 2610.39489v1

Interactive llm agents improve web accessibility error detection accuracy

InterA11y: Interactive Web Accessibility Error Detection with LLM Agents

Commercial implications: For digital accessibility service providers: Enables sales of improved accessibility auditing services with higher accuracy and efficiency to clients.

Thu 29 OctHuman-Computer Interaction
The gist
Websites need to be accessible to everyone, but current tools that check accessibility often miss tricky interactive problems or give too many false alarms. The authors created InterA11y, a system that uses detailed guidelines and interactive tools to help large language models test websites more carefully. This approach helps the system find more real accessibility errors while keeping false alarms low. InterA11y works well across different models and uses less computing power when checking websites.
Open → 2610.39488v1

Ultra low power underwater camera extends coral reef monitoring time

CoraLive: Ultra-Low-Power Underwater Vision for Long-term Coral Reef Monitoring

Commercial implications: For environmental monitoring device manufacturers: Enables production of long-lasting underwater cameras with optimized data processing for customers engaged in environmental monitoring.

Thu 29 OctNetworking and Internet Architecture
The gist
Protecting coral reefs requires watching them for a long time, but existing cameras run out of battery fast or need a lot of human help. The authors show a new underwater camera system that uses less power by taking lots of simple pictures but only a few detailed ones. It sends less data to shore by saving only what’s important, letting computers there do the heavy work. This way, the camera can last months much longer while still capturing useful information about the reefs.
Open → 2610.39472v1

RoFiGR improves item recommendations by refining coarse routing with collaborative signals

RoFiGR: Coarse Routing, Fine Refinement via Level-Adaptive Distillation in Multimodal Generative Recommendation

Commercial implications: For e-commerce platform developers: Enables improved personalized product suggestions that can increase sales and user engagement on online retail websites.

Thu 29 OctInformation Retrieval
The gist
Recommendation systems try to suggest items by predicting codes that represent each item step by step. Making mistakes early on can exclude good choices later. The authors present RoFiGR, which uses a two-stage approach: it first makes broad category decisions using information about user preferences and item similarities, then refines predictions carefully. This method improves recommendation quality on datasets like Instruments, Arts, and Games. The authors also show how training and decoding strategies affect the performance.
Open → 2610.39470v1

PathCowork trains AI by watching how pathologists work slides

PathCowork: Learning from real-world pathology practice to build AI copilots

Commercial implications: For medical device developers: The paper enables building AI pathology copilot products that incorporate expert visual and procedural data, differentiating them from standard AI diagnostics.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Medical AI usually learns only from the final answers doctors give, not how they examine slides to reach those answers. The authors created PathCowork, a system that records pathologists’ eye movements, spoken explanations, and navigation through slides to teach AI how to better interpret medical images. They collected detailed data from many pathologists reviewing microscope slides and used it to build AI agents that understand visual cues and reasoning steps. These AI helpers performed better and faster than existing models on several organ-specific tasks, showing promise for learning directly from doctors’ real-world work.
Open → 2610.39441v1

Simple method adds clear word emphasis to text to speech voices

EmpDPO: Enabling Emphasis Control in TTS via Direct Preference Optimization

Commercial implications: For audiobook producers: Allows selling more natural and expressive audiobook voices with fine-grained emphasis control.

Thu 29 OctSound
The gist
Some words need extra emphasis when spoken to show meaning beyond just the text. This paper shows how to make a speech program emphasize words better, even if it wasn’t trained with emphasis before. The authors use two training steps with a tool that detects emphasis automatically, so no one has to label examples by hand. Their improved program puts the right emphasis on words while still sounding natural.
Open → 2610.39427v1

Safe multi-objective reinforcement learning with zero constraint violations

Safe Multi-Objective Reinforcement Learning: Zero Constraint Violation and Primal-Averaging Convergence

Commercial implications: For autonomous vehicle engineers: Enables production-ready self-driving car systems that rigorously guarantee safety during development and deployment.

Thu 29 OctMachine LearningArtificial Intelligence
The gist
Balancing multiple goals safely is a big challenge when teaching computers to make decisions over time. The authors developed new methods that guarantee no safety rules are broken while the computer learns and explores. Their methods also find good trade-offs between goals, even when some preferences aren’t known at the start. This helps ensure safe and effective learning in complex environments where making mistakes can be costly or dangerous.
Open → 2610.39409v1

Efficient object detection speeds up fire smoke monitoring on devices

YOLOF-MLD: A Multi-Level Distillation Framework for Efficient Object Detection

Commercial implications: For smart home device makers: Enables safer, faster home safety products with advanced fire and smoke detection powered by lightweight neural networks.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Detecting objects quickly and accurately on small devices like Raspberry Pi is important for things like spotting fire or smoke. The authors show how replacing a complex part of a detection system with a simpler one can make it run faster but usually hurts accuracy. They fixed this by using a clever training method that teaches the simpler system to mimic the detailed one at multiple levels. Their method runs much faster on a Raspberry Pi with almost the same accuracy, making it better for real-world safety devices.
Open → 2610.39408v1

Utility-WAM improves robot actions by fixing future-related errors

Utility-WAM: Correcting Future-Induced Action Changes

Commercial implications: For manufacturing automation teams: Enables more reliable high-precision robot automation products in manufacturing lines by reducing failures from future-conditioned control.

Thu 29 OctRobotics
The gist
Robots can try to improve their actions by imagining what might happen next using visual predictions, but sometimes this makes them do worse on precise tasks. The authors found that adding this future prediction mostly helps with long tasks but can cause failures on delicate operations. They created Utility-WAM, a method that corrects these harmful changes while keeping the benefits for longer activities. Their system also reduces slowdowns from extra video processing by skipping it when it’s not needed.
Open → 2610.39367v1

Diffusion model improves same person face replacement in photos

PicturePerfect: Diffusion-based Same-identity Facial Replacement

Commercial implications: For photo editing software developers: Enables face replacement features to be sold as part of commercial photo editing tools that deliver more realistic and identity-preserving edits.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Replacing a person's face in a photo with another image of the same person is tricky, especially when the photos have different lighting or expressions. The authors developed a method that uses a special type of AI called a diffusion model to swap faces in pictures more naturally and realistically. They built a big dataset of photos showing the same person with different faces to help train and test their method. Their results look better and keep the person's identity more consistent than previous techniques.
Open → 2610.39344v1

MemoryFusion improves video fusion with long term memory and smoothing

MemoryFusion: Cross-Temporal Memory Learning for Multimodal Video Fusion

Commercial implications: For video editors: Enables creation of products offering advanced video fusion and stabilization features for professional video production.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Multimodal videos combine information from different sources but existing methods focus only on short-term frame interactions, missing important long-term details. The authors propose MemoryFusion, a two-part system that remembers both recent and distant frames to create better combined video content. Their method also smooths videos over time to reduce flickering and improve stability. Tests show MemoryFusion produces clearer images and more consistent videos than prior methods.
Open → 2610.39322v1

TaxoMFM improves gut microbiome analysis using taxonomic hierarchy

TaxoMFM: A Taxonomy-Aware Foundation Model for the Human Gut Microbiome

Commercial implications: For biotechnology companies: Enables creation of more accurate microbiome-based diagnostics for diseases like colorectal cancer and diabetes by exploiting microbial family structures.

Thu 29 OctMachine Learning
The gist
Gut microbiome studies often treat bacteria as unrelated groups, missing the natural family tree of microbes. The authors designed TaxoMFM, a model that understands these family connections to better learn patterns in gut bacteria data. They created a large dataset and showed that TaxoMFM predicts diseases and traits related to gut health with good accuracy, even for groups it was never trained on. The model also keeps consistent results when the bacteria database changes, making it more reliable for research and health applications.
Open → 2610.39316v1

Reinforcement learning improves multi image editing consistency

DeCoRL: Decomposed Consistency Reinforcement Learning for Multi-Image Composition

Commercial implications: For graphic design teams: Enables advanced commercial image editing software to reliably handle multiple-image compositions, improving product quality and user control.

Thu 29 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Editing images using several reference pictures at once is hard because keeping everything consistent across images is tricky. The paper presents a new approach called DeCoRL that teaches a model to focus on each object separately when combining images, so it keeps the details right. It uses a special reward system that checks object-level accuracy and helps the model learn better. They also created a new benchmark to test how well models can combine from two to six images, showing improved results.
Open → 2610.39265v1

Online reinforcement learning improves ios app navigation agents

Online Dense Reinforcement Learning with Planning-Grounding Agent for UI Navigation on Live IOS Apps

Commercial implications: For robotic process automation providers: This paper enables development of advanced automation products that function reliably on iOS and Android by training agents online.

Thu 29 OctArtificial Intelligence
The gist
Navigating apps on smartphones using AI is hard because big models are slow and smaller models often fail when things don’t go as planned. The authors designed a system that trains AI agents to learn by interacting directly with real iOS apps. They split the AI into two parts: one that plans steps and another that understands the app screens. They also use a smart feedback system to help the AI learn better at each step and introduced a new way to estimate values that makes learning more stable. Their approach trains agents that work better and faster than previous methods, even when tested on other devices and operating systems.
Open → 2610.39264v1

Personalized item views improve sequential recommendations over long histories

PRISM: Personalized Item Representations in User History

Commercial implications: For ecommerce platform teams: Enables product recommendation features with personalized item views that can boost sales and engagement on ecommerce platforms.

Thu 29 OctInformation RetrievalArtificial Intelligence
The gist
Recommendation systems usually treat items the same way for every user, giving each item a single representation. This paper presents a way to make item representations personalized based on each user's unique relationship with the item. The authors build a system called PRISM that creates these personalized embeddings from user history, improving how well the system can predict what a user wants next. Their tests on a large real-world dataset show better results as more user history is included.
Open → 2610.39175v1

VEvo-Harness improves self-updating AI for precise image creation

VEvo-Harness: Causally-Guarded Self-Evolving Harness for Agentic Image Generation

Commercial implications: For automated content generation services: It makes selling advanced image creation services with self-improving AI possible by ensuring reliable adaptation to user demands.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Creating images that perfectly match detailed user requests is difficult because current AI systems don't learn well from their mistakes and repairs. This paper introduces VEvo-Harness, a system that helps AI image generators learn continuously and improve over time by carefully tracking mistakes, verifying fixes, and storing useful skills. It uses a smart process to find the real cause of problems, tests possible fixes before applying them, and chooses the best way to repair images. The authors show that this approach works better than existing AI tools and matches some commercial image generators.
Open → 2610.39158v1

Federated learning boosts large model customization in web3

DICE: Decentralized Incentive-driven Federated Learning for Efficient & Scalable Customization of Pretrained Models in Web3

Commercial implications: For ai service providers: Enables offering secure, privacy-preserving AI customization services to decentralized user bases, differentiating product features.

Thu 29 OctCryptography and SecurityDistributed, Parallel, and Cluster ComputingMachine Learning
The gist
Customizing big AI models for specific tasks usually needs lots of data, which many Web3 apps have. The authors present a new method called DICE that helps many users customize these large models safely and efficiently without needing powerful computers. It lets users work on parts of a model rather than the whole, includes checks to keep the model accurate, and uses a special voting system to add security when many people join. Their tests show this method works well for both vision and language AI models.
Open → 2610.39132v1

Consumer embeddings improve ad ranking for food delivery marketplaces

PALATE: Pretrained Autoregressive LATent Embeddings

Commercial implications: For digital marketing teams: Enables improved personalized advertising products that increase marketplace monetization.

Thu 29 OctInformation RetrievalMachine Learning
The gist
Personalizing ads for consumers on local-commerce platforms is hard because each person interacts with only a few stores, making their history sparse. The authors built a model that predicts what stores a person might engage with next by learning from their past clicks and purchases. This model produces a single summary vector representing a person's preferences, which can then be used to show better ads across many surfaces. Their approach improved ad ranking accuracy offline and boosted revenue and conversions in live tests. Key innovations include training the model to focus more on new businesses, a two-step training process, and a carefully chosen set of negative examples.
Open → 2610.39090v1

How speech models differ on low resource African languages

How Architecture, Typology, and Adaptation Strategies Shape ASR for Low-Resource African Languages: A Controlled Empirical Study

Commercial implications: For language technology product teams: Enables commercial speech recognition services for previously underserved African languages with improved accuracy and efficiency.

Wed 28 OctComputation and LanguageSound
The gist
Creating speech recognition systems for African languages with limited data is hard. The authors compared three different speech AI models on five low-resource African languages and one medium-resource language. They found that some models handle data and language differences better than others, and that the amount and type of data impact accuracy. They also learned that adjusting models to match specific topics helps some systems but hurts others. Finally, languages’ unique characteristics affect how often models make mistakes.
Open → 2610.39004v1

AI peer agent seen as supportive partner in math learning

Exploring AI Peer in Collaborative Mathematics Learning

Commercial implications: For educational technology developers: Enables development of AI-driven math tutoring platforms that feel more like supportive peers, improving user engagement and learning outcomes.

Wed 28 OctHuman-Computer Interaction
The gist
It can be hard for people to see an AI as a true partner when solving math problems together. This paper looks at what makes an AI feel more like a teammate. The authors found that people trust the AI more when it acts like a real person—making mistakes, sharing control, and thinking out loud—rather than just giving correct answers or instructions. These insights help design AI that can better support collaborative learning.
Open → 2610.38990v1

Critical thinking improves with structured use of generative AI tools

When Does GenAI Use Support Critical Thinking? A Cross-Cultural Structural Account of Transactive Memory with GenAI

Commercial implications: For educational technology developers: Enables educational software companies to build smarter tutoring systems that promote deeper engagement with AI-generated content.

Wed 28 OctHuman-Computer Interaction
The gist
People wonder if using generative AI helps them think better or not. This study looked at college students in the US and Türkiye and found that just using AI isn't enough to improve critical thinking. Instead, how people work together with the AI — by sharing tasks and carefully rechecking its answers — is what matters. This pattern was mostly the same across the two cultures studied.
Open → 2610.38910v1

Improved code decoding methods reduce alphabet size significantly

Improved LCL Equivalences and Curve Decoding Alphabet Sizes

Commercial implications: For blockchain developers: Enables smaller and faster SNARK-based verifications, improving blockchain scalability and user experience.

Wed 28 OctInformation TheoryCryptography and Security
The gist
This work improves understanding of certain properties of error-correcting codes, which help computers fix errors in data. The authors build on earlier research to show that several complex code types behave similarly when decoding curves, a method important for verifying data efficiently. They also reduce how large the set of symbols (alphabet) needs to be for these codes to work well, making them more practical. This helps improve techniques used in very efficient proofs for checking computations.
Open → 2610.38862v1

Generative flow networks improve diverse financial crisis scenario simulation

Discovery of Diverse and Realistic Financial Tail-Risk Using Generative Flow Networks

Commercial implications: For financial software developers: Enables the development of advanced, market-ready financial risk management software offering more comprehensive scenario coverage.

Wed 28 OctMachine Learning
The gist
Predicting rare but impactful financial events is difficult because available methods either focus too narrowly or explore inefficiently. The authors introduce GRID, a new way to generate many different and realistic potential financial crisis scenarios using a special kind of sampler called a Generative Flow Network. By sampling proportionally to risk instead of just maximizing it, GRID can better explore a wide range of severe event possibilities even when the risk estimates are uncertain. Tests on real financial crises show that GRID creates more diverse and realistic risky scenarios than other approaches.
Open → 2610.38838v1

Vocal cord localization improved through local point tracking method

Beyond Region Memory: Local Point Tracking for Vocal Cord Localization in Video Laryngoscopy

Commercial implications: For medical device developers: Enables development of more reliable emergency intubation tools with better vocal cord visualization, improving patient outcomes.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
Finding the vocal cords accurately during emergency video laryngoscopy is difficult because the tissues move quickly and instruments can block the view. The authors propose a new way to track points in the video frames to keep detecting the vocal cords even when visibility gets tricky. Their method first finds key points on the vocal cords, follows these points through the video, and then adjusts the detected area to match the vocal cords more tightly. Tests on multiple datasets show this approach works better than previous methods, especially in keeping up with fast movements and improving boundary accuracy.
Open → 2610.38830v1

Visualization design techniques to match viewer thinking and goals

Crossing the Chasm: Bridging Visual Augmentations and Designer Intent

Commercial implications: For visualization software developers: Enables creation of smarter design software that helps users create effective augmented visualizations, increasing market value and usability.

Wed 28 OctHuman-Computer Interaction
The gist
Visualizations often use extras like notes or animations to guide what people pay attention to and understand. The authors found that existing rules mainly help with showing data but not with these extras aimed at sharing bigger meanings or ideas. They created a framework that connects the designer's communication goals with how viewers think, showing how different techniques like annotations or animations help achieve these goals. This helps make clearer visual stories and could enable tools to better support designers in the future.
Open → 2610.38782v1

Lightweight model achieves broad medical image segmentation with 211k parameters

A 211K Parameter Foundation Model: Genetic Prompting in Neural Cellular Automata

Commercial implications: For medical device developers: Enables creation of affordable medical imaging products for low-resource healthcare settings by reducing AI model complexity dramatically.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
Medical image segmentation is important for diagnosing many illnesses, but typically requires large computer models that need powerful equipment. The authors created a very small model called Foundational-NCA that can handle many different segmentation tasks without retraining. It uses a special technique combining cellular automata and genetic prompting to adapt dynamically to various tasks. This small model performs well on both known and new tasks, making it easier to use in places with limited computing resources.
Open → 2610.38776v1

Bayesian inquiry improves accuracy in multimodal diagnostic inference

Auditable Hierarchical Inference with Multimodal Evidence and Bayesian Active Inquiry

Commercial implications: For medical device developers: Enables development of smarter diagnostic tools that adapt questioning strategies for better accuracy, attractive to healthcare technology companies.

Wed 28 OctMachine Learning
The gist
Most systems that try to identify medical conditions using multiple signals assume all information is available right away, but real clinical checks happen step by step. The authors created a fake test setup that separates guessing the diagnosis from asking additional questions. Their system uses a smart questioning method to narrow down possibilities more accurately than fixed question orders, although it sometimes asks more questions. These results show a repeatable way to test such systems but do not prove they work in real medical settings.
Open → 2610.38774v1

Language model test tool generates many reasoning questions automatically

Deogen: Automatic Benchmark Generation for Reasoning over Policies

Commercial implications: For compliance software developers: Enables selling enhanced compliance testing platforms for regulated industries that require rigorous rule-following by AI models.

Wed 28 OctArtificial IntelligenceMachine Learning
The gist
It can be hard and expensive to create tests that check if AI models understand and follow complex rules correctly. The authors created a tool called DeoGen that automatically makes lots of these rule-following test questions by turning policies into logic trees. This way, many test cases are generated quickly, including ones where not enough information is given so the model should not guess an answer. They found that while AI models often get the right answer when enough information is available, they struggle when information is missing, even when told to admit uncertainty.
Open → 2610.38687v1

Financial assistants often answer wrong questions with right numbers

Right Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries

Commercial implications: For financial technology developers: Enables creation of smarter financial chatbots and advisory tools that reduce misinterpretation risks for end users.

Wed 28 OctArtificial IntelligenceComputation and Language
The gist
Financial assistants sometimes give a number that looks correct but actually answers a different question than the user intended. The authors show this happens when questions leave out important details, like which financial product or tax rule applies. They studied Indian mutual funds where details are well defined, finding that models often pick default or outdated assumptions without realizing it. This means even answers that match data can be wrong for the user’s situation. The authors offer ways to detect and control these mistakes.
Open → 2610.38631v1

ItemCompass helps find misplaced household objects with wearable AI

ItemCompass: Real-Time Edge-AI Wearable Vision for Tag-Free Domestic Object Retrieval

Commercial implications: For assistive technology developers: Enables production of assistive smart glasses marketed to aging populations needing support with daily object retrieval.

Wed 28 OctHuman-Computer Interaction
The gist
Losing everyday things at home, especially for older people, is frustrating. The authors created ItemCompass, a smart pair of glasses that recognize and remember objects without needing tags. It listens to your voice and guides you to lost items using augmented reality. Their tests show the system is accurate and trusted by users while protecting privacy.
Open → 2610.38629v1

Smart wheelchair navigates safely with limited and faulty user commands

A Multi-Layer Navigation System for Smart Wheelchairs with a Limited Input Interface

Commercial implications: For robotics product developers: Enables commercial development of more reliable smart wheelchairs that better interpret user intent in real settings.

Wed 28 OctRobotics
The gist
Smart wheelchairs help people move around safely but can struggle if the user’s commands are unclear or if the environment is new. The authors created a new navigation system that can understand limited or wrong user commands and safely guide the wheelchair through unknown places. Their system combines sensors, smart planning, and vision techniques to detect obstacles, plan paths, and identify targets. They tested it on a real wheelchair with 10 people and showed it works well even when commands are imperfect.
Open → 2610.38569v1

Ai assistants learn personalized behavior from feedback without retraining

Persona Learning from Feedback for Personalized AI Assistants

Commercial implications: For ai product developers: Enables creation of customizable AI assistants with reusable behavior modules for consumer or enterprise applications.

Wed 28 OctMachine LearningArtificial IntelligenceComputation and Language
The gist
Personalizing AI assistants to fit different users or tasks can be tricky because it’s not always clear what the AI should change or keep in its responses. The authors introduce a new way to teach AI assistants personalized behaviors using user feedback without needing to retrain the entire system. Their method creates reusable behavior modules that can be selected automatically based on the conversation, improving how well the assistant matches user preferences. Tests show this approach works better than older methods that use fixed personalities or prompting alone.
Open → 2610.38561v1

OmniQuad robot moves easily with wheels and legs combined

OmniQuad: a Wheeled-Legged Hybrid Robot with Omnidirectional Wheels

Commercial implications: For indoor logistics teams: Enables selling advanced delivery or material handling robots that combine fast indoor movement with obstacle navigation.

Wed 28 OctRobotics
The gist
Robots that walk on legs are good at handling rough ground but move slowly indoors, while wheeled robots are fast on smooth floors but struggle with obstacles. The authors created OmniQuad, a robot that combines legs with special wheels that let it move in any direction. This design helps the robot move quickly inside buildings and also handle steps and uneven terrain outside. They built the robot, designed its controls, and tested it in different environments.
Open → 2610.38502v1

Continuous space time video super resolution improves clarity and smoothness

Residual Motion Splatting for Continuous Space-Time Video Super-Resolution

Commercial implications: For video streaming services: Enables products that upscale streaming content with better quality and responsiveness, appealing to media service providers.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
Low-resolution, choppy videos can be made clearer and smoother by increasing their details and frame rates. The authors developed a method that predicts how each pixel moves over time with high precision, even between original frames. Their technique combines multiple blurry video frames onto a sharper grid while handling missing parts or overlaps, making the process faster and resulting in better quality. This approach merges frame interpolation and resolution enhancement into one step and works well on standard benchmarks.
Open → 2610.38499v1

Generative AI helps game creators collaborate but raises trust concerns

Generative AI as a Mediator in Creator Collaboration: Challenges, Design Opportunities, and Concerns

Commercial implications: For creative agencies: Enables creation of AI-driven teamwork platforms sold to agencies to enhance multi-disciplinary project outcomes.

Wed 28 OctHuman-Computer Interaction
The gist
When different creative experts work together, it can be hard for them to understand each other fully. The authors studied how generative AI can help writers and artists make games by acting as a go-between that aligns their ideas and reduces conflicts. However, they found that AI might also cause people to talk less, trust each other less, and change how they work together. The authors suggest ways AI tools can support teamwork without hurting creativity or trust.
Open → 2610.38488v1

StreamStudio lets creators shape streaming videos while they generate

StreamStudio: Exploring Video Authoring with Streaming Autoregressive Generation

Commercial implications: For video producers: Enables a new product for professional video editing software that provides live generative assistance during content creation.

Wed 28 OctHuman-Computer Interaction
The gist
Making videos with AI usually means describing what you want, waiting for the whole video, then fixing problems afterward. The authors present StreamStudio, a tool that lets users shape and change the video while it’s being made, like steering a story as it happens. This means creators can try different ideas continuously and make changes at any point without waiting for the full video first. The system is flexible and can work with various video generation models, aiming to make video creation more interactive and controllable.
Open → 2610.38467v1

IntReal dataset and benchmark improve text and image generation together

IntReal: A Real-World Dataset and Benchmark for Multimodal Interleaved Generation

Commercial implications: For digital content creators: This enables creation of advanced content generation software offering superior multimodal interleaved outputs for marketing and media companies.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
People often read or view content where text and images are mixed together naturally, but computers struggle to make this kind of mixed content well. The authors created IntReal, including a big real-world dataset and a detailed way to test how good computer models are at making mixed text and images. They showed their tests agree well with human opinions and can spot where different models do better or worse. They also found that training models on just text or just images can help them get better at making mixed content, which could save effort when building these systems.
Open → 2610.38425v1

Governance-aware federated learning improves trust in shared AI training

Governance-Aware Federated Learning for Multi-Agent AI Systems: A Policy-Driven Framework for Trustworthy Collaborative AI

Commercial implications: For cloud service operators: Enables cloud providers to sell federated learning services with enhanced security and compliance features.

Wed 28 OctArtificial Intelligence
The gist
Federated learning lets multiple users train an AI together without sharing their private data, but it can be hard to know if everyone is following the rules or being fair. The authors propose a new system called Governance-Aware Federated Learning (GAFL) that checks evidence of good behavior, fairness, and trust for each participant during training. They tested GAFL with 10 participants and found it was able to detect bad or dishonest actions effectively, though this added some processing time and made fairness harder to balance. This work shows how AI collaboration can be monitored better to build more trustworthy shared models.
Open → 2610.38411v1

LLMs predict medical exam question difficulty from their own responses

LLMs as Judges and Examinees for Medical Item Difficulty Prediction

Commercial implications: For educational assessment vendors: Enables development of commercial difficulty prediction tools for secure exam items using deployable open-weight LLMs.

Wed 28 OctComputation and LanguageArtificial Intelligence
The gist
Understanding how hard medical exam questions are is tricky just from reading their text. The authors studied large language models (LLMs) by making them act like test graders and also pretend to be students taking the tests. These roles let the models produce different clues about how difficult each question is. Combining these clues gave better predictions than previous methods on two big medical exam datasets. Their approach can be used locally on secure or unreleased exam questions.
Open → 2610.38402v1

Quadruped robots learn new whole-body skills from failure feedback

ELEVATE: Agentic Controller Acquisition from Task Failure for Quadrupedal Manipulation

Commercial implications: For warehouse automation teams: Enables sale of more versatile quadruped robots capable of handling complex material handling tasks autonomously.

Wed 28 OctRobotics
The gist
Quadruped robots that can move and manipulate objects often need new skills to complete complicated tasks but don’t always know what skills they lack. The authors present ELEVATE, a system where the robot notices when it fails at a task, then automatically creates and improves new controllers to fix that problem using learning techniques. These new skills are added to the robot’s library, allowing it to solve harder problems. Their approach was tested in simulations and on a real robot without extra training, achieving high success on placing and retrieving objects.
Open → 2610.38388v1

Irish text-to-speech improved using pronunciation guides and small data

Pronunciation-Supervised Text-to-Speech for Irish (Gaeilge)

Commercial implications: For language technology companies: Enables improved commercial Irish speech products by leveraging pronunciation data to enhance pretrained models with limited audio.

Wed 28 OctComputation and Language
The gist
Making computers speak Irish well is hard because there is little recorded speech for training. The authors show it helps to use written pronunciation guides from dictionaries to teach a powerful speech model how to pronounce Irish words. They fine-tuned a large pre-existing model with a small amount of Irish speech and improved accuracy by checking pronunciations against dictionary rules. This approach also measures where spoken data is missing, like rare dialects.
Open → 2610.38384v1

Generative recommendation improves advertising revenue and efficiency

Towards Reliable, Scalable, and Value-Aligned Generative Recommendation for Large-Scale Advertising

Commercial implications: For online advertising engineers: Enables improved ad delivery platforms that increase revenue and meet performance requirements for large-scale digital advertisers.

Wed 28 OctInformation RetrievalMachine Learning
The gist
Online advertising platforms need to show personalized ads that users are likely to engage with while keeping the process fast. The authors improve an existing system that generates ad identifiers by making it more reliable, scaling its capacity efficiently, and better aligning recommendations with business value. Their improved system, tested on a large platform, increased revenue by about 3% while meeting strict speed and throughput requirements. This work helps balance the trade-offs between effectiveness and efficiency in ad recommendation.
Open → 2610.38378v1

Tripolarized antennas unlock more spatial streams in wireless communication

Is Dual Polarization Always Optimal? Unlocking New Spatial Degrees of Freedom

Commercial implications: For smart home device manufacturers: Enables selling higher-capacity wireless devices by leveraging tripolarized antennas to boost local data transmission efficiency.

Wed 28 OctInformation Theory
The gist
Wireless systems often use two polarizations of antennas because it works well when signals travel far and behave like flat waves. This paper shows that when devices are closer, a special three-polarization antenna can send more independent signals, improving data capacity. The authors built and tested such antennas indoors and outdoors and found they could send up to three separate data streams instead of just two. This could help future wireless systems send more information at once.
Open → 2610.38377v1

SparseTokenGS improves fast low-dose 3D X-ray image reconstruction quality

SparseTokenGS: Gaussian-based Sparse-View CBCT Reconstruction via Reusable Token Pool

Commercial implications: For medical imaging device developers: Enables commercial CBCT products with faster, lower-radiation scans selling improved clinical imaging for dental and intervention use.

Wed 28 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Taking 3D X-ray images quickly with fewer pictures helps reduce radiation but causes blurry or distorted images. The authors introduce SparseTokenGS, a new method that uses a shared set of data 'tokens' to keep information from the entire scan while reconstructing the image, leading to clearer results than before. This approach works in one step without needing extra adjustments for each scan. It has been tested on different medical and non-medical objects, showing better image quality and useful results for dental treatments like root canals.
Open → 2610.38360v1

Volleyball 3D reconstruction improves player and ball tracking accuracy

Volley3D: Unified 3D Full-Sequence Reconstruction from Monocular Broadcasts

Commercial implications: For sports ar/vr developers: Enables production of realistic 3D VR sports replays from casual broadcasts, expanding market offerings in sports entertainment.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
Figuring out the full 3D positions of players and the ball in volleyball games from just one TV camera is very hard because the camera moves and volleyball moves in complex ways like high jumps and quick ball hits. The authors created a new method that does not need 3D training data and uses physics ideas to correct errors caused by the camera’s perspective and reconstruct the ball’s path. Their system breaks down the ball’s movement into smaller parts and matches collisions to improve accuracy. Tests show their approach tracks players and the ball in 3D much better than simpler methods that just guess from 2D images.
Open → 2610.38358v1

Llm agents benchmarked on forecasting quarterly earnings per share

EPSBench: Benchmarking LLM Agents for Earnings Forecasting

Commercial implications: For investment software developers: EPSBench enables creating improved AI-driven earnings forecasting tools for investment platforms, offering competitive advantage.

Wed 28 OctComputation and LanguageArtificial Intelligence
The gist
Forecasting a company's quarterly earnings involves understanding many financial details and how they add up to reported profits. The authors built a test called EPSBench where AI models try to predict quarterly earnings by reading official documents and news about companies. They compare how well these AI models do against professional human analysts. The results show some models perform about as well as analysts, but prediction intervals often underestimate actual earnings. Most errors come from estimating a key income figure called operating income.
Open → 2610.38293v1

Vendster measures forecast bias effects on AI buyer vendor deals

Vendster: An Evaluation Arena to Investigate How Forecasting Tool Quality Affects LLM Agent Behaviors in Buyer-Vendor Negotiation

Commercial implications: For financial forecasting providers: Enables development of specialized forecasting services tailored for automated negotiation platforms to increase commercial value.

Wed 28 OctComputer Science and Game TheoryArtificial Intelligence
The gist
Forecasts usually get judged by how accurate they are, but what really matters is how useful they are for making decisions. The authors created Vendster, a tool that tests how the kind of forecast, especially its bias, changes negotiation results between two AI agents acting as a buyer and a vendor. They found that certain biases, like consistently underpredicting demand, can actually help the agents achieve better sales targets, more than just accuracy scores suggest. Vendster lets anyone test different forecast types to see their impact on decisions.
Open → 2610.38292v1

Cross lingual summarization improves Marathi to English text summaries

Cross Lingual Abstractive Summarization for Marathi

Commercial implications: For multilingual content platforms: Enables content platforms to offer better cross-lingual summarization features, attracting more users and advertisers.

Wed 28 OctComputation and LanguageArtificial Intelligence
The gist
Summarizing text in different languages is hard, especially for Indian languages like Marathi. The authors worked on making summaries that translate Marathi text into English summaries using advanced machine learning models. By combining existing Marathi data and a smaller translated dataset, they improved the quality of these summaries better than previous methods. Their work mainly focused on translating Marathi to English summaries but discussed the reverse direction as well.
Open → 2610.38267v1

FHE data processing boosted by smart in-storage acceleration approach

Rosemary: Efficient and Adaptable In-Storage Acceleration for Fully Homomorphic Encryption

Commercial implications: For enterprise storage designers: This paper enables building SSDs with embedded FHE engines for secure client data processing, creating new product opportunities for storage hardware vendors.

Wed 28 OctCryptography and SecurityHardware Architecture
The gist
Fully Homomorphic Encryption (FHE) lets computers work on encrypted data, but it is very slow and costly because moving data between storage and processors takes up most of the time. The authors show that current systems spend 71% to 85% of time just moving data instead of computing. They created Rosemary, a new system that processes encrypted data directly inside storage devices using special hardware and clever software tools. This approach speeds up FHE tasks like encrypted AI and database queries by over 180 times compared to standard hardware and uses much less energy.
Open → 2610.38262v1

Keepace AI specializes large language model for Chinese exercise guidance

Keepace.ai: A Technical Report on a Large Language Model for Sports and Health

Commercial implications: For health app developers: Enables development of advanced fitness apps offering tailored exercise guidance based on a specialized AI model for Chinese speakers.

Wed 28 OctComputation and Language
The gist
Providing safe and personalized exercise advice using large AI models is challenging because there is little good training data and no clear way to judge accuracy in sports and health topics. The authors created Keepace.ai by continuing to train a large model specifically for Chinese language exercise plans, sports knowledge questions, and personal data analysis. Internal tests by experts showed Keepace.ai gives safer, more accurate, and more complete workout advice than general models. This report explains how Keepace.ai was built, tested, and evaluated for safety and scientific reliability.
Open → 2610.38249v1

Silhouette analysis helps assess frailty in older adults from walking videos

Silhouette-Based Gaitomics for Frailty Assessment in Older Adults

Commercial implications: For healthcare technology developers: Enables commercial development of non-invasive frailty assessment devices or software for clinics and home use.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
Frailty in older adults affects health and mobility but can be hard to measure. The authors developed a method that looks at simple body outlines—called silhouettes—captured from walking videos to find patterns linked to frailty. Their approach picks out key features that describe body shape and movement over time, connecting these features to how frail a person is. This method can separate frail individuals reasonably well from others, though it struggles more with less severe cases. More studies with larger groups are needed to confirm its usefulness.
Open → 2610.38223v1

Building damage assessment improved by reducing neighbourhood shortcut errors

Neighbourhood shortcuts in building damage assessment from very-high-resolution satellite imagery: diagnosis and a building-evidence plug-in

Commercial implications: For satellite image analytics companies: This plug-in can be sold as a software upgrade improving model reliability for clients in insurance or emergency services.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
It is hard to tell if computer models that look at satellite images to judge building damage after disasters are really assessing each building or just guessing based on nearby buildings. The authors found that many models often rely too much on the damage status of neighboring buildings, missing isolated damaged ones or incorrectly flagging intact buildings in damaged areas. They created an add-on that helps models look at individual buildings more carefully without retraining the whole system. This add-on improves accuracy, fixing many mistakes caused by relying too much on neighbors and spotting isolated damaged buildings better.
Open → 2610.38162v1

StepControl generates full-body motion from just foot contact timing

StepControl: Long-Horizon Human Motion Generation from Foot-Contact Schedules

Commercial implications: For game developers: Enables game studios to offer customizable character movements based on footstep schedules, enhancing animation tools for interactive gameplay.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
People often control animated human motion by telling a computer what action to do or where to move. This paper looks at controlling motion by just telling when each foot touches the ground. The authors created StepControl, a model that generates full-body movement using only these foot contact timings, without needing other details like body positions. Their model follows the foot timing well and can handle changes like stops or limps, showing that just foot contact timing can guide realistic motion.
Open → 2610.38153v1

Ai agent optimizes warehouse layout to speed up picking and throughput

AI-Driven Digital Twin Framework for Warehouse Logistics Automation Optimization

Commercial implications: For supply chain software vendors: This paper enables development of AI-enhanced warehouse software products that optimize slotting using real-time data and simulations.

Wed 28 OctArtificial Intelligence
The gist
Warehouses traditionally arrange products in fixed spots, which is often inefficient when conditions change. The authors created a smart computer agent that learns from real-time data and adjusts where items are placed in a warehouse to make picking faster and increase overall productivity. They tested this in a detailed warehouse simulator with forklifts and found their method reduces the time to find items by almost 30 percent and makes the system work 35 percent faster. This approach uses a digital model of the warehouse that talks to the agent and real assets seamlessly.
Open → 2610.38101v1

Nemotron 3.5 models improve AI reasoning and multimodal tasks efficiently

Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI

Commercial implications: For software engineering teams: Enables advanced AI assistant products for software developers with improved reasoning and interactive tool use.

Wed 28 OctComputation and LanguageArtificial IntelligenceMachine Learning
The gist
Large AI models can struggle to think through complicated problems or handle long pieces of text and different types of information like images and videos. The paper introduces two versions of the Nemotron 3.5 model, designed to solve these problems more efficiently while working with multiple types of data. One big model, called Super, can handle text, images, and videos and is trained to improve reasoning and software tasks. The smaller model, Lightning, focuses on text and is easier to deploy. The authors show these models perform well on tasks involving Indian languages, software engineering, and speed-heavy AI use cases.
Open → 2610.38095v1

Cash flow predictions improve small business credit decisions

CADENCE: Distributional Next-Transaction Forecasting for Business Cash-Flow Underwriting

Commercial implications: For financial institutions: Allows lenders to offer better credit decisions based on enhanced cash flow forecasts, improving financial product offerings for small businesses.

Tue 27 OctMachine Learning
The gist
Predicting when and how much money a small business will receive regularly can be tricky because payments are often unpredictable and vary in timing. The authors created CADENCE, a compact AI model that treats each transaction like a word in a sentence to forecast the next payment amount and timing more accurately. It uses knowledge about the calendar and descriptions of transactions to improve predictions, especially for businesses with limited payment history. These better predictions help lenders make smarter credit decisions by understanding cash flow patterns more clearly.
Open → 2610.38084v1

Monocular depth estimation works well on lunar terrain images

Efficacy of Monocular Depth Estimation in Lunar Environments

Commercial implications: For virtual reality content creators: Enables production of commercial lunar VR environments using fewer sensors, reducing complexity and cost.

Tue 27 OctRoboticsComputer Vision and Pattern Recognition
The gist
Getting a sense of depth from just one camera image is a tough challenge, especially on the moon where conditions are very different from Earth. This paper looks at how well modern computer techniques for estimating depth work when applied to pictures of the moon's surface. The authors tested different methods using special Earth-based lunar-like images and real Apollo 17 photos. They found that some methods can still create good 3D maps of the moon’s uneven surface, even though these tools were mostly trained on Earth scenes.
Open → 2610.38080v1

Conformal prediction methods improve delivery accuracy and cost savings

Conformal Predictive Distributions for Contextual Order Fulfillment Optimization

Commercial implications: For supply chain software developers: Enables development of delivery optimization software with superior prediction accuracy and cost benefits for retailers.

Tue 27 OctMachine Learning
The gist
Deciding where and how to ship orders before knowing exact delivery times is tricky because delivery delays vary. The authors use new prediction methods called conformal predictive distributions to better estimate delivery times and their uncertainties. These improved estimates help reduce costs and late deliveries for online orders. Tests on millions of real orders and a delivery simulator show notable improvements in accuracy and cost.
Open → 2610.38046v1

Sentiment aware prompts improve mental health chatbot responses

PromptSense: Sentiment-Aware Mental Health Chatbot with Dynamic Prompting

Commercial implications: For mental health app developers: Enables development of improved mental health chatbot products with better user engagement and safety, appealing to healthcare providers and app markets.

Tue 27 OctArtificial Intelligence
The gist
Mental health chatbots often miss emotional cues and safety concerns, which can lead to less helpful or even harmful conversations. This paper shows a way to make chatbots better by first analyzing the user's emotions and depression signs, then adjusting the chatbot's prompts accordingly. The authors tested different methods and found that their approach improves how empathetic, clear, and safe the chatbot's replies are. This could help make mental health chatbots more supportive and trustworthy.
Open → 2610.38044v1

Tennis event detection struggles beyond broadcast camera views

Can Event Detectors Leave the Broadcast Camera? Evidence from Tennis

Commercial implications: For video technology developers: This paper enables new video analysis products for tennis broadcasters and content creators by improving event detection across various viewpoints.

Tue 27 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Detecting events like ball bounces and racket hits in tennis videos is easy when using broadcast footage, but much harder with videos from other angles like phones or court-level cameras. The authors created a benchmark with annotations from different viewpoints and tested several detection methods. They found that models trained on broadcast videos perform poorly on other views, while a new method, RallyArb, works better on some camera angles but has timing issues with hits. This work highlights the challenges of detecting tennis events outside traditional broadcast views.
Open → 2610.38042v1

Signature analysis of wearable data detects brain cancer progression

Detecting Brain Cancer Progression Through Signatures of Wearable Accelerometry Data

Commercial implications: For digital health device developers: Enables development of wearable-based monitoring products targeting neuro-oncology patients for continuous disease tracking.

Tue 27 OctMachine Learning
The gist
Brain cancer patients often experience worsening symptoms over time, but traditional scans to monitor this are expensive and infrequent. This paper shows that data from wrist-worn devices tracking movement can help identify when the disease is getting worse. The authors convert walking data into patterns and use a mathematical technique called rough path signatures to spot changes specific to each patient. This method could lead to cheaper, continuous monitoring of brain cancer progression outside the hospital.
Open → 2610.38021v1

Multimodal attacks find weak spots in language AI understanding images

Where Can We Fool the Model's Mind? Response-Guided Semantic Probing and Adaptive Routing for Transferable Attacks on Black-Box MLLMs

Commercial implications: For content moderation teams: Enables tools that automatically flag or block adversarially altered images targeting AI moderators, protecting platform integrity.

Tue 27 OctComputer Vision and Pattern Recognition
The gist
Some AI models combine language and images to understand the world, but they can be tricked by small changes in pictures. The authors created a method called VISTA that finds exactly where these AI models are most likely to fail by watching how the models answer questions about images. This method works without needing to look inside the models and can fool even commercial AI systems. Their approach can both guess weak points using open AI models and also improve attacks by asking the target model questions.
Open → 2610.38013v1

Smart home technologies need adjustment for autistic children’s needs

Making Smart Homes Fit for Autistic Children: Family Work and Monitored Independence

Commercial implications: For smart home product designers: Enables development of specialized smart home products that better serve autistic families, expanding market reach and user satisfaction.

Tue 27 OctHuman-Computer InteractionArtificial Intelligence
The gist
Smart home devices are common in many homes, but they don’t always work well for children with autism and their families. The authors studied how families use, change, and manage these technologies to help their autistic children while facing challenges like sensory overload and unclear device behavior. They also found that parents balance safety, privacy, and independence by monitoring their children remotely. The authors introduce new ideas to better understand how technology fits into family life and suggest ways to improve device design.
Open → 2610.38008v1