Papers for

e-commerce platform developers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Multimodal recommendation improves product suggestions using images and reviews

A Technique to Model Product Information and User Reviews (MPIUR)

Abstract: Modern e-commerce platforms naturally provide three complementary modalities: textual item descriptions, item images, and user reviews. However, existing multimodal recommendation methods often omit reviews or do not integrate all three modalities within a unified framework. In this paper, we propose MPIUR, a multimodal graph-based recommendation model that encodes reviews in heterogeneous graphs, aligns image and text representations, and adapts collaborative--semantic alignment to item interaction density. Experiments on the Amazon Reviews 2023 Musical Instruments and Baby categories show that MPIUR achieves the best global nDCG@10 and Recall@10 among the evaluated methods on both datasets. It also obtains the best nDCG@10 and Recall@10 on sparse items, indicating that jointly exploiting reviews and multimodal item content is particularly useful when interaction histories are limited.

Mon 9 NovArtificial IntelligenceInformation Retrieval
The gist
Online stores show product descriptions, images, and user reviews, but many recommendation tools don’t use all three together. The authors created a method called MPIUR that combines item images, texts, and reviews in a graph model to better suggest products. They tested it on Amazon’s musical instruments and baby product reviews and found it gave better top recommendations, especially for items with fewer past purchases. This shows using all types of product info helps recommend less popular items more effectively.
Open → 2611.13361v1

User sequence modeling improves search product ad ranking

Query- and Candidate-Aware Long User Sequence Modeling for Sponsored Product Ranking in Search

Abstract: User sequence modeling is widely used in recommendation and advertising. Search ranking presents an additional challenge: the current query expresses immediate intent, while a candidate product determines which parts of a user's long-term history are relevant. We present a production sequence framework for sponsored-product ranking that augments recent behavior with two retrieval channels over long-term history: candidate-nearest-neighbor retrieval and query-nearest-neighbor retrieval. The selected actions are encoded together with dedicated query and candidate context tokens, while candidate-aware early fusion injects the candidate into every history token. Offline ablations show that dual retrieval improves major ranking metrics relative to the production baseline. In an online A/B test on sponsored-product search traffic, the model increased CTR, attributed conversion rate, and orders per impression.

Sun 8 NovMachine Learning
The gist
Online shopping search rankings try to show ads that match what you're looking for. The paper's team improved these rankings by not only using your recent activity but also by looking back into your long-term history related to both the current search and the ad candidate. They created a model that mixes context from both the search term and the candidate product to better guess which ads you'll click and buy. Their tests showed this model did better in predicting clicks and purchases than older methods.
Open → 2611.11675v1

Recommendation system captures changing user preferences with context aware attention

An Adaptive, Context-Aware, and Stacked Attention Network-Based Recommendation System to Capture Users' Temporal Preference

Abstract: Recommendation systems have become more important since the widespread use of the Internet. The large amount of information means that it is difficult for users to discover they really need. However, users' preferences can change over time because of the age and the impact of social networks. Recommendation systems must capture users' preferences and recommend suitable items, which is a difficult challenge. This study proposes a novel recommendation system that adapts to the changing preferences of users. By learning and adapting to users' changing preferences better recommendations are provided. This study uses context factors as additional information to model users' preferences more precisely. For the proposed recommendation system, the context is based on users' interaction items within each defined time interval. This study proposes two novel attention mechanisms for the recommendation system. The contextual item attention module captures contextual information, changing pattern in users' preference and the importance of items. The multi-head attention module extends the diversity of users' preferences and adapts to changing preferences. The recommendation performance is improved using additional item's temporal information to model the contextual item's representation. Experiments compare the proposed algorithm with several state-of-the-art recommendation methods using three real-world datasets. The experimental results demonstrate that the proposed context-aware recommendation model outperforms traditional methods and demonstrate the effectiveness with which contextual information is captured by the attention mechanism.

Sun 8 NovInformation RetrievalMachine Learning
The gist
People’s tastes for things they like can change over time, making it hard for recommendation systems to suggest what they want. The authors designed a system that looks at what users interact with over different times and learns from the changing patterns. It uses special attention methods to focus on important items and different user interests. Their method worked better than others on real-world data sets.
Open → 2611.11278v1

Fair decoding method improves fairness in language model recommendations

Conformal Fair Decoding for Large Language Model Recommenders

Abstract: Large language models (LLMs) can generate personalized recommendations without task-specific training, but changing a protected attribute while holding interaction history fixed can alter the semantic content of recommended items. We study fairness control when model weights and full-vocabulary logits are unavailable. We propose Fair- Rec, an item-level decoding framework that uses an offline-calibrated semantic-disparity threshold and controls candidate selection as recommendation lists are generated. A hard variant filters candidates against the threshold; a soft variant trades off candidate likelihood and disparity while adapting a penalty using observed violations. Under an explicit uniform enforceability assumption, we establish a pathwise O(1/T) upper bound on the excess empirical violation rate of the soft decoder. Experiments on MovieLens-1M and Amazon Movies & TV using Qwen3-8B and Llama3.1-8B-Instruct show that the soft variant lowers measured counterfactual disparity relative to zero- =shot recommendation, UP5, and FACTER, while retaining competitive HR@10 and NDCG@10.

Fri 6 NovArtificial Intelligence
The gist
Recommendations made by large language models can unfairly change when a user’s protected attribute (like gender) is altered, even if their past interactions stay the same. The authors propose a new method called FairRec that controls which items these models recommend to reduce unfair differences without needing full access to the language model internals. Their approach adjusts recommendations by filtering or penalizing items based on their fairness, and experiments showed it keeps recommendation quality while improving fairness compared to previous methods.
Open → 2611.10285v1

Platform competition can help users when credibility is assessed reliably

Toward User-Centered Agentic Recommendation: When Does Platform Competition Benefit Users?

Abstract: Conventional recommender systems operate within platform-specific catalogs, limiting access to potentially valuable resources across platforms. We propose user-centered agentic recommendation, a paradigm in which user agents compare recommendations from competing platform agents, and formalize their interaction as a repeated game. Platforms strategically modify descriptions without changing intrinsic item value, while each user agent assesses platform credibility using feedback from consumed items only. Our analysis shows that expanded resource access need not improve user welfare: strategic misrepresentation and limited verification can offset the benefits of platform competition. We quantify when platform credibility assessment allows users to benefit from broader access despite strategic exaggeration. Our analysis identifies resource and reporting conditions that guarantee higher average welfare than the original platform with an interaction budget linear in platform count. Experiments on synthetic and Amazon-grounded markets show that credibility assessment can improve finite-horizon utility, while its benefits depend on market size and mechanism parameters.

Tue 3 NovArtificial Intelligence
The gist
Recommender systems usually suggest items from only one platform, which limits user options. The authors study a new approach where user agents compare recommendations from multiple competing platforms and evaluate their trustworthiness based solely on feedback from used items. They find that having access to more platforms does not always help users, because platforms may exaggerate descriptions to attract users, and users have limited ability to verify these claims. However, when users can accurately judge platform credibility over repeated interactions, they can achieve better satisfaction from wider choices. Their experiments show this benefit depends on market size and how users assess credibility.
Open → 2611.04203v1

Recommendation algorithms improve by coordinating design choices and experiments

RecRSI: Mechanism-Guided Discovery of Recommendation Algorithms

Abstract: Improving recommendation algorithms for the Web requires coordinating interacting design choices and learning from experiments. We present RecRSI, an LLM-based research framework that uses mechanism hypotheses to organize algorithm discovery. Each hypothesis links a proposed mechanism change to a reference design, a rationale, and an expected observation. MechSpace derives implementable primitives and composition rules from the task source; Research Line uses this space and prior evidence to develop and test hypotheses; Mechanism Feedback carries outcomes and their design context into subsequent research. This organization makes each experiment inform which mechanisms to develop, revise, or combine next. Across four initial recommenders and four datasets, RecRSI yields average relative improvements of 8.48% in NDCG@5 and 6.88% in HR@5 over the initial algorithms, and 2.60% and 2.20%, respectively, over the strongest competing result for each setting and metric. Component and process studies link these gains to coordinated mechanism choices and evidence-guided follow-up. Our code is publicly available at https://github.com/DynamisG210/RecRSI.

Mon 2 NovInformation RetrievalArtificial Intelligence
The gist
Recommendation algorithms help decide what content or products to show you online, but making them better is tricky because many design choices interact with each other. The authors created a framework called RecRSI that uses clear ideas about how changes to recommendation designs should work and then tests those ideas step-by-step. This approach helps the algorithms improve by learning which changes actually work well together. They tested their method on several algorithms and data sets and found it consistently improved recommendation quality.
Open → 2611.02487v1

Recommendation system improves ratings by combining explicit and hidden criteria

Multicriteria Recommendation System by Leveraging Predefined, Implicit, and Undefined Criteria

Abstract: Data sparsity, a long-standing problem in the recommendation system, occurs when the user-item utility matrix is sparse. For years, many approaches have been experimented with, including utilizing user's textual reviews and multiple criteria ratings. In a review-based recommendation system, user preferences and item attributes are learned from reviews, while in multi-criteria recommendation systems, criteria ratings are used for the recommendation task. Multi-criteria are domain-specific pre-defined explicit criteria and users are bound to provide feedback only on these criteria with discrete values. On the contrary, users are free to express their opinions through textual reviews, which can cover multiple perspectives or aspects. It is possible that these pre-defined, domain-specific explicit criteria are not enough to fully capture the multi-faceted user preferences. To this end, we propose a novel multi-criteria recommendation model that utilizes predefined, implicit, and undefined criteria. We use a semantic similarity-based sentence clustering method to identify the predefined and implicit criteria and a sentiment analyzer to estimate their ratings. The criteria that are extracted from the review and are aligned with predefined criteria are referred to as implicit criteria. A sentence is considered as expressing opinions on an undefined criterion if the similarity score between this sentence and all the predefined criteria is lower than a predefined threshold. Ratings are computed for each extracted implicit criterion and the undefined criterion based on the review content. Finally, we use all three types of criteria and an aggregation model to make the final rating prediction for the recommendation system. Our proposed method demonstrates the superiority compared to several baselines on TripAdvisor and Beer Advocate datasets.

Sun 1 NovInformation Retrieval
The gist
Recommendation systems often struggle when users provide very little feedback. Traditional systems either use fixed criteria or rely on text reviews separately. This paper presents a method that mixes pre-set criteria, hidden ones found by grouping similar opinions in reviews, and entirely new criteria that don’t match existing categories. The system calculates ratings from all these types and combines them to better predict user preferences. The authors tested this approach on travel and beer review sites and showed it works better than some existing methods.
Open → 2611.02040v1

RoFiGR improves item recommendations by refining coarse routing with collaborative signals

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

Abstract: Generative recommenders predict an item's semantic identifier (SID) one code at a time. Early codes select broad branches, so a routing error can exclude relevant items before fine-grained prediction begins. In multimodal recommendation, these coarse decisions must reflect collaborative preferences as well as item semantics. We introduce RoFiGR, a two-tree recommender that uses collaborative information in both SID construction and generation. It aligns cumulative L1-L2 code vectors with frozen collaborative item embeddings and explicitly encodes item distance and source level in the serialized history. During generation, a sequential teacher supervises legal branches. Fixed coarse-level weights maintain routing supervision, while a bounded adaptive residual allocates additional weight according to teacher-student demand. On Instruments, Arts, and Games, RoFiGR improves all five ranking metrics over the evaluated external baselines. Five-seed paired tests with Holm correction support the main NDCG@10 improvements. Comparisons with alternative collaborative supervision objectives and tokenizers support the proposed design. Matched task and decoding analyses distinguish the effects of training exposure, stream fusion, and candidate budgets.

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

Generative recommendation improves advertising revenue and efficiency

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

Abstract: Generative recommendation (GR) reformulates item retrieval as autoregressive generation of discrete semantic identifiers (SIDs), offering a unified paradigm for personalized recommendation. GR4AD, as an industrial generative recommendation framework, has greatly improved advertising revenue while enabling efficient real-time serving in Kuaishou.However, continued deployment has exposed three limitations: deterministic SID assignment remains sensitive to semantic boundaries; conventional backbone scaling provides diminishing revenue gains under strict serving constraints; and sequence-level value optimization provides limited credit assignment to intermediate SID decisions, causing likelihood-based beam search to prematurely prune potentially high-value candidates. We present GR4AD-X, which addresses these limitations through reliable SID construction, serving-efficient capacity scaling, and unified value alignment. Specifically, GR4AD-X improves SID reliability through Multi-Path assignment. It expands model capacity along multiple complementary dimensions rather than relying solely on wider or deeper backbones, achieving a better effectiveness-efficiency trade-off under strict serving constraints. Finally, Unified Value-Guided Generation and Discrimination (UniVGD) aligns generation with commercial value through prefix-level reward optimization, value-guided beam pruning, and discriminative reranking within a unified model. Large-scale online A/B tests on the Kuaishou advertising platform demonstrate that GR4AD-X achieves a cumulative revenue improvement of 3.07% over the deployed GR4AD baseline while satisfying strict latency and throughput constraints.

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

Improving item recommendations by modeling semantic and collaborative relations

REGER: Modeling inter-item Relations with Semantic IDs for Generative Recommendation

Abstract: Personalized recommendation depends on learning semantic and collaborative relations among items from online content and user interactions. Generative recommendation (GR) offers a promising paradigm by autoregressively generating discrete semantic IDs (SIDs) of items. However, randomly initialized SID token embeddings and GR models trained from scratch do not directly inherit the semantic relations inherent in items, while inter-item collaborative relations are learned largely implicitly through token-level attention and next-token supervision. To address this limitation, we propose REGER, a unified framework that explicitly incorporates inter-item relations into semantic pretraining, history encoding, and SID generation. We pretrain the GR model through generative semantic retrieval, enabling it to inherit inter-item semantic similarity by learning to generate semantic neighbors' SIDs. We combine global attention with local attention aligned with SID boundaries to capture historical context and directed transitions between adjacent items. We further develop a theoretical framework for hierarchical preference decomposition over the SID prefix tree, proving an exact factorization of collaborative preference distributions and a weighted decomposition of item-level KL divergence into prefix-conditional KL divergences. Building on this analysis, we distill a discriminative teacher's full conditional distributions over all legal next tokens at ground-truth prefixes along the target SID path, jointly optimizing this objective with cross-entropy. Extensive experiments on real-world datasets demonstrate consistent improvements over existing GR baselines and support the effectiveness of the proposed preference distillation.

Tue 27 OctInformation RetrievalArtificial Intelligence
The gist
Personalized recommendations often rely on understanding how items relate based on their meaning and user behavior. The authors found that current generative recommendation methods start without capturing these relationships well. They created a new approach called REGER that teaches the system to recognize and use connections between items, both in their meaning and in how users interact with them. Their method uses smarter ways to look at item history and includes a way to break down user preferences more precisely, helping the system suggest better next items. Experiments show this approach works better than earlier methods.
Open → 2610.37584v1

Sequential recommendation improved with cross-user behavior and token-level credit

RPSD: Retrieval-Privileged On-Policy Self-Distillation with Optimal Transport

Abstract: Sequential recommendation predicts a user's next interaction from historical behavior. Recent LLM-based methods recast this task as generative recommendation and increasingly adopt reinforcement learning such as GRPO for post-training. However, these methods suffer from two limitations, namely credit-blind optimization, where a sequence-level reward is shared across all generated tokens, and collaborative-blind supervision, where each training signal is derived from a single user and ignores behavioral patterns shared across similar users. To address both issues, we propose Retrieval-Privileged on-policy Self-Distillation (RPSD). For credit-blind optimization, we introduce on-policy self-distillation (OPSD), in which a teacher with privileged knowledge re-evaluates the student's on-policy response, and the token-level confidence difference between the student and teacher reshapes the shared sequence-level advantage into token-specific credit. However, the privileged knowledge in vanilla OPSD is still derived from a single user, leaving collaborative-blind supervision unresolved. We therefore ground the teacher in collaborative behavioral evidence from similar users. Specifically, we use temporally aware optimal transport (OT) to compare interaction trajectories across users by jointly considering item similarity and temporal structure, and further apply diversity-aware selection to retain complementary trajectories as privileged evidence. In this way, on-policy self-distillation provides fine-grained token-level credit, while OT enriches the teacher with structured cross-user evidence. To the best of our knowledge, this is the first work to bring on-policy self-distillation to sequential recommendation and to use cross-user behavior as its privileged signal. Extensive experiments across multiple real-world datasets demonstrate that RPSD achieves state-of-the-art performance.

Tue 27 OctArtificial Intelligence
The gist
Sequential recommendation tries to predict what a user will want next based on their past choices. The authors point out two problems with current methods: they give the same credit to all parts of a sequence of suggestions and only learn from one user's data at a time. They propose a new technique called RPSD that uses a teacher model to give detailed feedback on each recommendation step and includes information from similar users’ histories. To do this, they measure similarities between user behaviors over time with a method called optimal transport. Their approach improves the accuracy of recommendations on real data.
Open → 2610.37281v1

New method improves image and long text matching by balancing key and background details

Mitigating Saliency Collapse: Robust Saliency-Aware Long-Text Image-Text Alignment

Abstract: Long-text image-text alignment is essential for fine-grained scene understanding, yet existing methods mainly focus on extending context length or improving local alignment. We identify an overlooked failure mode, saliency collapse, where models over-rely on text that features visual prominence while underutilizing non-salient context. Consequently, retrieval performance degrades sharply when salient content is reduced, despite that the non-salient context remains informative. To address this, we propose RoSA, a saliency-aware finetuning framework that decomposes image-text pairs into complementary salient and contextual views, enabling joint learning of prominent and contextual semantics. We propose a modality-independent decomposition strategy to combine object-level visual saliency estimation with LLM-based text decomposition to bridge vision and language. We employ auxiliary objectives to explicitly align these views, ensuring robust representation with both salient and contextual features. RoSA greatly improves long-text retrieval and demonstrates strong robustness to saliency collapse across benchmarks. Moreover, our auxiliary objectives are plug-and-play, bringing consistent gains to existing tuning methods. Code and checkpoints are available at: https://github.com/intelligolabs/RoSA

Thu 22 OctArtificial Intelligence
The gist
Matching images to long descriptions can be tricky because models often focus too much on the most obvious parts and ignore the less noticeable but still important details. This causes problems when those obvious parts change or disappear. The authors found this issue, which they call saliency collapse, and created a method named RoSA that teaches models to pay attention to both the standout and background information. RoSA uses a clever way to split images and text into important and contextual parts and aligns these parts together. This results in better match accuracy and more reliable performance even when obvious cues are missing.
Open → 2610.31005v1

UserArena benchmark advances long-term shopping agent simulations

UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents

Abstract: Agentic recommendation pursues shopper goals through sequences of context-dependent decisions. Developing and evaluating such policies requires faithful recommendation world models that capture both the interactive shopping environment and users' behavioral responses over long interactions. Existing resources rarely provide long-horizon, fine-grained real-user trajectories or support trajectory-level evaluation and offline validation against production A/B results. We introduce UserArena, a real-world e-commerce interactive benchmark containing 64,729 session trajectories from 11,010 users, spanning 566,102 items and 13 action types. UserArena pairs these trajectories with an interactive environment and supports offline validation against outcomes from a controlled production A/B test, enabling five-level evaluation from output validity and local decisions to sequential behavior, personalization, and outcomes. To improve simulator fidelity over these long-horizon interactions, we further propose On-Policy Trajectory Distillation (OPTD), which uses trajectory-level intent to provide dense teacher targets on student-generated rollouts. Experiments show that OPTD achieves the strongest overall fidelity among the evaluated simulators under UserArena's evaluation protocol and most closely reproduces the production policy effect in offline simulation.

Thu 22 OctInformation Retrieval
The gist
Shopping recommendation systems try to help buyers by making a series of decisions that guide purchases over time. The authors created UserArena, a large collection of real shopping sessions with lots of details, and a test environment where recommendation methods can be evaluated more realistically over long interactions. They also developed a new training approach called On-Policy Trajectory Distillation to improve how well simulators mimic real user behavior during long shopping sessions. Their experiments show this method best matches actual outcomes recorded in a production test, making simulated training more reliable.
Open → 2610.30932v1

Onelatent boosts recommendation speed and accuracy using latent reasoning

OneLatent: Latent Reasoning for Efficient Foundation Recommendation Models

Abstract: Large language models (LLMs) have demonstrated strong reasoning capabilities, motivating their use as the backbone of foundation recommendation models (FRMs). Existing methods enhance recommendations through explicit Chain-of-Thought (CoT) reasoning under a Think-then-Answer paradigm. However, explicit CoT incurs substantial inference overhead by generating lengthy reasoning traces and relies on manually designed templates that struggle to capture diverse, dynamic user interests. We propose OneLatent, an efficient latent reasoning framework that compresses explicit reasoning traces into several learnable latent tokens, enabling Latent-Reason-then-Answer inference without generating verbose traces. OneLatent first introduces Multi-View Adaptive CoT (MV-ACoT), which creates diverse, high-quality teacher-generated supervision by exploring user interests from multiple perspectives and automatically adapting reasoning complexity to each instance. Building on pretrained FRMs, it then uses a three-stage latent-token alignment paradigm to progressively internalize CoT traces into learnable latent tokens. Finally, a multistage curriculum-based post-training strategy activates latent-token reasoning for downstream recommendation tasks. Experiments on an industrial-scale Kuaishou dataset and the public Kuaishou LLM-Rec benchmark show that OneLatent consistently outperforms explicit CoT-based methods and traditional baselines. Compared with the Think and No-Think variants of FRMs, OneLatent improves SID@64 by 17.44% and 9.33%, respectively, while achieving over 17x higher online inference throughput. We further develop a production serving system for scalable, real-time FRM inference. An online A/B test in Kuaishou's local-services advertising scenario shows that deploying OneLatent with this system yields an estimated 9.6% revenue lift over strong online baselines, including OneRec and OneReason.

Thu 22 OctInformation Retrieval
The gist
Recommender systems often use large language models that think step-by-step to improve suggestions, but this takes a lot of time and effort. The authors designed OneLatent, which shrinks these step-by-step thought processes into a few internal tokens, making recommendations faster and still accurate. They created a new way to teach the model to adapt reasoning to each user's interests and progressively learn these shortcuts. Tests on a big real-world dataset showed OneLatent gives better recommendations and is much faster than existing methods. When put into production for online ads, it increased revenue by almost 10%.
Open → 2610.30675v1

Evolving semantic indexes improves search and recommendation quality

Evolving Semantic Indexes for Generative Retrieval

Abstract: Generative retrieval (GR) casts search and recommendation as autoregressive generation of discrete semantic identifiers (SIDs). SID design largely determines GR quality, but its discrete, interdependent components (SID construction, collision handling, decoding, and training) cannot be optimized end-to-end with gradients and remain largely hand-designed. We introduce a framework that improves these components through LLM-driven evolutionary program search, to our knowledge the first application of this approach to GR. A locally served, open-weight 31B model mutates executable GR components, so no proprietary LLM API is used, and a 1M-parameter proxy decoder provides fast feedback in place of repeatedly training the 1B target generator. Across 21 SID programs on NQ320K, proxy and target Recall@100 show strong rank correlation (Spearman $ρ= 0.90$). Guided by this proxy, evolution improves the seed GR on NQ320K, MSMARCO300K, and a multilingual e-commerce dataset under a fixed training recipe. On NQ320K, jointly evolving SID construction and collision handling improves Recall@1 from 55.8% to 59.3% and Recall@100 from 85.8% to 90.6%. A component study shows that proxy reliability depends on what is evolved: improvements transfer reliably for the semantic index but not for the generator's training objective or the full pipeline. LLM-driven evolutionary search is thus a practical tool for GR design, provided proxy reliability is established for the chosen search space.

Tue 20 OctArtificial IntelligenceInformation RetrievalMachine Learning
The gist
Search engines and recommendation systems can work better by representing information with special codes called semantic identifiers. These codes are tricky to design because their parts interact, and improving them by hand is hard. The authors used a clever process where an AI program tries out many small changes to these codes and learns which ones work best. This method helped improve search accuracy on several datasets without relying on costly AI services. It shows that evolving parts of the system with fast testing can make search smarter.
Open → 2610.28568v1

E-commerce AI agents struggle to recover from silent tool errors

RITE: Evaluating End-to-End Resilience to Silent Tool Errors in E-Commerce Agentic Systems

Abstract: Tool-using LLM agents increasingly answer e-commerce queries through multi-step tool interactions. The expected output of a tool call can shift between agent development and execution, allowing a plausible but incorrect value to be returned without an explicit failure signal. These silent errors may propagate through subsequent decisions into final answers. To study agents' resilience across this end-to-end process, we formulate Silent-error Resilience Task, an interactive task that evaluates whether they complete the original query after a silent error corrupts an intermediate state. We introduce RITE, a benchmark built from human-authored industrial workflows and production-log-derived inputs. To enable controlled and reproducible evaluation, we use a semantic error taxonomy to inject errors into fixed workflow prefixes, freeze recorded tool behavior, and let agents continue freely. It contains 167 matched units across nine e-commerce workflows, yielding 501 silent-error, correct-value, and explicit-failure instances. Across 10 recent LLMs, silent errors roughly halve task success and are consistently harder than explicit failures. Recognition, use of an available verification path, and integration of the resulting evidence form successive bottlenecks. We also propose a light State Checker as an intervention to test whether resilience can be improved.

Tue 20 OctArtificial Intelligence
The gist
In e-commerce AI systems that use multiple tools step-by-step, mistakes can quietly slip in when a tool returns a believable but wrong answer without signaling a failure. These hidden errors can affect later decisions and final results. The authors created a test called RITE to check how well AI agents handle these silent mistakes and whether they can still complete tasks correctly. They found that these silent errors significantly reduce success rates and are harder to detect compared to clear failures. They also tested a simple checking tool to help improve error recovery.
Open → 2610.27777v1

WenSight decision model improves e commerce rule outcomes accuracy

WenSight-Decision-1 for Direct Decisions under E-commerce Rules

Abstract: E-commerce decisions require applying natural-language rules to the available case evidence. We introduce WenSight-Decision-1, a shared model that scores supplied options directly from rules and case information for six operational tasks, with examples spanning six languages. We fine-tune Qwen3.5-9B with reference options as supervision, using the same option scores for training and inference. On 1,440 synthetic operational cases reserved for final evaluation, accuracy rises from 61.60% for Qwen3.5-9B to 92.92% for WenSight-Decision-1 under identical inputs and scoring. Joint fine-tuning improves all six tasks, bringing heterogeneous operational judgments under a common decision interface.

Mon 19 OctComputation and Language
The gist
Making decisions in online shopping often relies on understanding complex rules written in plain language. The authors created WenSight-Decision-1, a model that directly compares possible choices against these rules using examples in six different languages. It learns by using scores as examples and performs better than its predecessor, achieving much higher accuracy on test cases. The model also works well across several different e-commerce decision-making tasks using a unified approach.
Open → 2610.26444v1

Recommendation system improves cold-start item suggestions using dual-branch graph learning

Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation

Abstract: Graph Collaborative Filtering models have achieved remarkable success in recommendation systems. However, they inherently suffer from a catastrophic representation collapse in strict cold-start scenarios due to the topological isolation of unobserved items. While integrating multi-modal semantic features offers a potential remedy, directly mapping isolated semantic attributes into a globally topology-driven collaborative space frequently precipitates representation degradation and optimization conflicts. To bridge this fundamental structural divide, we propose DualGCN, an innovative Dual-Branch Graph Knowledge Distillation Framework. Operating under a decoupled two-stage training paradigm, DualGCN first constructs a robust collaborative teacher network where item representations are anchored by a topology-smoothed semantic prior, effectively reshaping the latent manifold. Subsequently, a semantic student branch employs a residual mapping mechanism and a magnitude-agnostic cosine distillation objective to achieve high-fidelity cross-modal alignment without disrupting established collaborative signals. Furthermore, we introduce a dynamic norm reconstruction mechanism to rigorously calibrate the physical scales of novel and warm items during global retrieval. Extensive experiments across three real-world benchmark datasets demonstrate that DualGCN significantly outperforms state-of-the-art baselines in strict cold-start scenarios while comprehensively preserving recommendation accuracy for highly active items.

Wed 14 OctInformation RetrievalMachine Learning
The gist
Recommendation systems struggle to suggest new items that have no user interaction history, resulting in poor performance for these 'cold-start' items. The authors propose DualGCN, a two-part method that first creates strong item representations using network information and semantic features, then carefully aligns semantic information without disturbing the network-based knowledge. This approach helps the system understand new items better without losing accuracy on popular items. Experiments show their method works well on real datasets.
Open → 2610.19578v1

Simple SimHash tokens match complex methods in item recommendations

Rethinking Semantic ID Construction for Generative Recommendation: SimHash with Parallel Decoding and Semantic Alignment

Abstract: Semantic ID-based generative recommendation represents each item as a sequence of discrete tokens, enabling structured modeling of item semantics. A critical challenge is constructing semantic IDs that are both semantically expressive and computationally efficient. While recent approaches favor complex learned quantization, simple hashing-based methods such as SimHash are widely regarded as fundamentally inferior. In this work, we challenge this consensus by showing that the apparent performance gap does not stem from inherent limitations of hashing, but rather from a structural mismatch with autoregressive decoding, coupled with the inevitable information loss during rigid discretization. Based on this insight, we propose FLASH, a two-stage framework that revitalizes training-free SimHash tokenization through parallel decoding and explicit semantic alignment. Despite its simplicity, FLASH achieves state-of-the-art performance across multiple datasets without requiring any tokenizer training, while exhibiting stronger generalization in cold-start scenarios. Notably, we demonstrate that semantic alignment acts as a universally effective mechanism across diverse paradigms. Our findings suggest that, with compatible decoding and semantic grounding, simple and efficient tokenizers can achieve performance comparable to complex learned counterparts in generative recommendation. Our code is available at https://github.com/KevinC2015/Flash.

Mon 5 OctInformation Retrieval
The gist
Recommender systems often represent items using sequences of tokens that capture their meanings. Normally, creating these tokens involves complex training, but this paper shows that a simple hashing technique called SimHash can perform just as well if combined with certain decoding and alignment steps. The authors introduce FLASH, a method that uses SimHash without any training and improves recommendation results, especially when suggesting new or rare items. This challenges the belief that only complex learned methods can work well for such tasks.
Open → 2610.07402v1

E-commerce video reasoning boosted by active evidence seeking model

Beyond Single Videos: Benchmarking and Active Evidence Seeking for E-Commerce Cross-Video Reasoning

Abstract: E-commerce videos are information-dense and frequently compared by consumers evaluating products and merchants assessing marketing strategies. However, existing multimodal models mainly focus on single-video understanding and have limited ability to compare information across videos. We introduce AdsCVR, the first e-commerce cross-video reasoning benchmark, containing 2,483 videos and 6,110 question-answer pairs across six reasoning dimensions. Cross- video reasoning requires models to locate fine-grained evidence among many redundant frames and integrate visual details, speech, and on-screen text. We therefore propose AdSeek, an agentic framework that dynamically selects visual and audio tools during multi-turn exploration, replacing static uniform sampling with active evidence acquisition. To address the sparse credit assignment of reinforcement learning, we develop an offline trajectory rectification mechanism that identifies reasoning errors and missing multimodal evidence in RL-generated trajectories. The corrected trajectories provide supervised fine-tuning signals that reduce biases learned during RL. This mechanism supports a rectified bootstrapping pipeline in which initial RL exposes reasoning bottlenecks, supervised fine-tuning corrects them, and a final RL stage further improves the policy. AdSeek achieves 74.30 percent accuracy on the AdsCVR test split, outperforming its Qwen3-VL-8B-Instruct backbone by 27.90 percentage points. It also generalizes to the open- domain CrossVid benchmark, demonstrating effective active evidence gathering.

Fri 2 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
People often compare products by watching many e-commerce videos, but current AI models mostly understand videos one by one and struggle to compare details across multiple videos. The authors created AdsCVR, a new test set with thousands of videos and questions to challenge AI to reason across videos using images, speech, and text. They also developed AdSeek, a smart system that actively picks when and where to look and listen in videos to gather the right clues, improving reasoning accuracy. This approach notably outperforms prior models and works well on other video reasoning tasks too.
Open → 2610.03099v1

Recommendation systems improve by combining semantic and collaborative signals

Neither Black nor White: Balancing Semantic and Collaborative Signals with Graph-Informed Semantic IDs (GrIS)

Abstract: Existing work on Semantic IDs (SIDs) for generative recommendation treats SID construction as a representation learning problem: encode items into a quantised latent space and read off codes. We argue this view is incidental. SID construction is, at heart, a recursive clustering problem, and once stated this way the natural object to cluster is a graph whose nodes carry semantic content and whose edges carry collaborative signal; SID assignment becomes a hierarchical graph partition. This reframing yields a unified framework, Graph-Informed Semantic IDs (GrIS), that subsumes prior approaches rather than displacing them. RQ-VAE and RQ-KMeans are recovered as the special case where the graph is empty, exposing content-only quantisation as one corner of a larger design space along two so-far-collapsed axes: graph construction and recursive partition algorithm. We explore two contrasting instantiations: RecDMoN, which performs hierarchical assignment via differentiable graph pooling, and RQ-GAE, which extends RQ-VAE with graph-aware item representations and a graph reconstruction objective. On multiple real-world datasets, GrIS consistently improves over CF-aware SOTA, with gains of up to +52\% Hit@10. Because graph construction and partition are explicit, separately configurable components, improvements on either axis can be combined and evaluated systematically.

Thu 1 OctArtificial IntelligenceInformation RetrievalMachine Learning
The gist
Generating good product or content recommendations can be tricky because you want to consider both what the item is about (its meaning) and how users interact with it. The authors argue that instead of just encoding items into categories based on meaning alone, it's better to think of items as nodes in a network that connects similar things by meaning and user behavior. They created a new way to group items on this network to make recommendations that better balance content and collaboration information. Their method leads to more accurate recommendations on real datasets compared to existing techniques.
Open → 2610.01533v1

Multilingual encoders struggle to produce consistent semantic codes across languages

Do Multilingual Encoders Produce Language-Consistent Semantic IDs?

Abstract: Semantic IDs (SIDs) compress item embeddings into discrete code sequences used in generative retrieval. We ask whether a multilingual encoder is sufficient for different-language renderings of the same product to receive language-consistent SIDs. Using Amazon ESCI listings rendered in English, Spanish, and Japanese, we test whether translations remain close to their English source, whether residual quantization is unusually sensitive to translation-induced movement, and whether multilingual or language-balanced quantizer fitting improves SID agreement. Multilingual E5 places translations measurably apart: under an English-heavy fit, a Japanese translation preserves the first SID code of its English counterpart in only 7.7% of cases, compared with 89.0% for an English rewording. Distance-matched product-directed controls produce nearly the same full-SID mismatch as translation, providing no evidence that the quantizer selectively amplifies language directions. Balancing the fitting mixture makes codebook use more uniform but further reduces cross-lingual prefix agreement: Spanish first-code consistency falls from 28.3% to 6.6%, while an English-only fit preserves it for 67.6% of Spanish translations. These results show that multilingual exposure and balanced codebook use alone do not guarantee language-consistent SIDs.

Thu 1 OctInformation RetrievalComputation and Language
The gist
Semantic IDs (SIDs) are special codes that represent product information in a compact way, used for searching and retrieving items. This paper examines whether a multilingual encoder can assign consistent SIDs to the same product described in different languages. The authors found that translations often get very different SIDs compared to the English original, meaning the system doesn’t reliably recognize them as the same item across languages. Adjusting training to balance languages or quantizer settings did not solve this issue, showing that more work is needed for language-consistent semantic coding.
Open → 2610.01139v1

Residual trajectory distillation improves generative retrieval accuracy

Residual Trajectory Distillation for Generative Retrieval

Abstract: Generative retrieval has emerged as a general retrieval paradigm, representing items with discrete Semantic IDs (SIDs) and retrieving them through autoregressive identifier generation. When SIDs are constructed with residual quantization (RQ), standard retrieval training supervises only the selected codes and discards the residual trajectories that produce them. The same hard code can nevertheless arise from different preferences over competing codewords, while the residual trajectory also contains information about subsequent quantization decisions. As a result, hard SID supervision collapses distinct quantization behaviors into identical targets and leaves information available during indexing unused in retrieval training. We introduce ResTD, a Residual Trajectory Distillation framework that transfers this discarded indexing information into retrieval training. Treating the frozen RQ indexer as a process teacher, it distills residual-induced codeword preferences into SID-decoding states. This supervision recovers distinctions hidden by hard assignments and allows earlier decoder states to capture information about subsequent quantization decisions before the corresponding SID suffix is generated. In this way, richer information from SID construction is incorporated into retrieval learning while preserving the original retrieval index and inference procedure. Experiments on multilingual e-commerce retrieval show consistent improvements over strong baselines and matched training controls. Controlled comparisons show that residual-derived targets outperform the tested codebook-only soft targets. Representation probes further show that future codebook preferences become more recoverable from earlier decoder states. ResTD can also be readily extended beyond retrieval to generative recommendation. Code is available at: https://github.com/Nevaeh7/iclr2027_ResTD.git.

Wed 30 SeptInformation Retrieval
The gist
Search systems sometimes represent items using compact digital codes and try to find them by predicting these codes step-by-step. Traditional methods only learn from the final chosen codes, ignoring helpful clues about how those codes were made. The authors introduce a new way to teach retrieval models to use these clues from the entire coding process, improving search performance. They tested this idea on multilingual online shopping searches and found it made the system better at retrieving relevant items.
Open → 2609.39319v1

VADER speeds up filtered vector searches with automatic recall control

VADER: Filtered Vector Search with Declarative Recall

Abstract: Approximate filtered vector search (FVS), a core operation in many data management tasks that combine structured data with vector embeddings, exhibits increased complexity due to the characteristics of filtering predicates. Each predicate is defined by selectivity (i.e., the fraction of vectors that satisfy the predicate) and correlation (i.e., the relationship between the filter and the vector space), which can significantly affect search difficulty even for the same query vector. This poses a key challenge for users aiming to integrate vector search with structured data, as efficient execution often requires extensive manual tuning of algorithm parameters. In this paper, we present VADER, the first approach that eliminates hyperparameter tuning by introducing declarative recall for approximate filtered vector search. With declarative recall, users specify a desired recall target, and VADER executes FVS queries to meet this target without requiring manual configuration. VADER achieves this by employing a filter-aware recall predictor that generalizes across varying selectivities and correlations without explicit tuning, and by performing early termination once the predicted recall reaches the user-defined target. Through extensive experimental evaluation, we show that VADER achieves near-optimal early termination, while providing significant speedups of up to 53% faster and improved result quality of 28% compared to the best-performing baseline.

Tue 29 SeptDatabases
The gist
Filtered vector search helps computers find related items based on both data details and numeric patterns, which is usually complicated and needs a lot of fine-tuning. The authors present VADER, a system that lets users set how complete they want their search results to be, and then automatically adjusts the search process to meet that goal without manual tweaking. VADER predicts how well the search is doing and stops early when the target is reached, making searches faster and better. Tests show that VADER is often quicker and provides higher quality results than previous methods.
Open → 2609.38053v1

Recommendation agents improve using screenshots and dynamic memory

ReMem: Rethinking Perception and Memory in Long-Context Recommendation Agents

Abstract: Recent Recommendation Agents (RecAgents) offer a promising alternative by shifting recommendation to an active, user-side paradigm, where generative agents autonomously perceive external platforms, reason over user preferences, and execute decisions. However, existing RecAgents still suffer from two critical limitations: brittle item perception based on noisy and heterogeneous item pages, and inefficient long-context reasoning over extended user histories and multi-step interaction traces. To address these challenges, we propose a novel recommendation agent framework, termed as ReMem, that combines OCR-based multimodal perception with time-evolving dynamic memory. Instead of parsing raw HTML, ReMem observes item pages through screenshots and extracts structured multimodal information via an OCR tool, enabling a more humanoid and platform-agnostic perception mechanism. To support long-horizon preference modeling, ReMem further introduces a chunk-wise sequential memory update strategy, where the agent selectively maintains a fixed-size memory of informative historical interactions while processing arbitrarily long contexts with linear inference complexity and bounded context length. This design allows the agent to preserve evolving user preferences without relying on external memory modules or disrupting the standard autoregressive generation process. To enhance the dynamic memory instruction, we further develop a multi-memory GRPO variant, which propagates the final-answer advantage to all intermediate conversations that contribute to the final response. Extensive experiments on three datasets demonstrate that ReMem consistently outperforms state-of-the-art baselines, achieving an average improvement of 5.16\% across three recommendation agent tasks, namely searching, ranking, and judging.

Tue 29 SeptArtificial IntelligenceInformation Retrieval
The gist
Recommendation agents help suggest items users might like, but existing ones have trouble understanding messy web pages and remembering long user histories efficiently. The authors propose ReMem, a new system that reads item information from screenshots using OCR, making it less dependent on web page structure and more adaptable. ReMem also uses a special memory method to keep track of user preferences over long times without slowing down. Tests show ReMem performs better than current top methods in recommending, ranking, and judging items.
Open → 2609.37311v1

Session recommendation gets faster and better without attention models

No Attention, No Problem: Rethinking Session-based Recommendation with Pure Convolution

Abstract: Session-based recommendation (SBR) predicts the next choice in a session by analyzing recent interactions. Transformer-based models are widely used because of their ability to capture long-range dependencies through self-attention mechanisms. In contrast, traditional convolutional models, although more efficient, are often limited by their weak global modeling capabilities and are losing ground in SBR tasks. In this work, we propose a Next-generation Pure Convolutional Framework (NextConvRec) for SBR tasks, aiming to balance efficiency and performance. NextConvRec uses a Structural and Positional Convolutional Encoder (SPCE) for preprocessing, combining learnable convolutional positional biases with session-level structural signals extracted through GCN layers. Its backbone convolutional module effectively expands the effective receptive field through depthwise convolutions and pointwise convolutions, enabling robust long-range preference modeling without attention mechanisms. Extensive experiments on 4 benchmark datasets show that NextConvRec outperforms several state-of-the-art baselines by around 1.73% on average, and reduces the average inference time per session by 16.7%. The convolutional architectures remain a promising direction for efficient and accurate session-based recommendations.

Mon 28 SeptInformation Retrieval
The gist
Session-based recommendation systems try to predict what you will choose next based on your recent actions. Usually, complex models that use attention mechanisms help these systems understand long-term connections but are slower. The authors show that a new convolution-based method can do this faster while keeping or even improving accuracy. This method combines positional signals and structural information without using attention, making it more efficient for real-time predictions.
Open → 2609.34802v1

Recommendation models improve generalization beyond one epoch training

Beyond One Epoch: Uncertainty-Weighted Sensitivity Regularization for Recommendation Models

Abstract: Recommendation models with sparse embeddings and a shared consumer often exhibit the one-epoch phenomenon: a second epoch lowers training loss while sharply degrading generalization. We present a view based on the violation of the prequential principle. On the first epoch, an example's label has not affected the embedding rows used to score it. On later epochs, those rows contain a displacement induced by the labels earlier update. This creates an incentive for the shared consumer to exploit this displacement in subsequent epochs, which fails to generalize. We call this self-influence asymmetry. Using an exact scalar model and local influence analysis, we connect this mismatch to the uncertainty in the embeddings and the consumers incentive to exploit it in subsequent epochs. We verify this hypothesis using an embedding-consumer-update interventions in deep recommendation models and propose uncertainty-weighted sensitivity regularization (UWSR) which counteracts this mismatch by augmenting the loss function to penalize the consumer for relying on uncertain embeddings. Unlike existing remedies, UWSR preserves the learned embeddings and across three benchmarks, four-epoch UWSR reduces test cross-entropy by 1.38%-6.78% and improves AUC by 0.0058-0.0231 relative to one-epoch training.

Mon 28 SeptMachine Learning
The gist
Recommendation systems often train quickly in one pass but get worse when trained longer, a problem called the one-epoch phenomenon. The paper explains this happens because the system unintentionally uses information from earlier training in a way that does not work well for new data. The authors show that this issue is linked to uncertainty in the model’s internal parts and propose a new training method called uncertainty-weighted sensitivity regularization (UWSR). UWSR helps the model avoid relying too much on uncertain parts, improving recommendation quality over multiple training passes.
Open → 2609.34083v1

Preserve and compose training improves zero shot composed image retrieval

Preserve-and-Compose Training for Composed Image Retrieval

Abstract: Composed image retrieval (CIR) aims to retrieve images that satisfy a user-specified modification while preserving relevant visual content from a reference image. Collecting target images for this purpose is costly, motivating zero-shot CIR methods that instead use target captions as supervision. However, target captions may omit source details that should be preserved. We therefore propose, Preserve-and-Compose Training, which complements target-caption supervision with visual evidence from the source image. PACT learns from image--text--text (ITT) triplets without target images or gallery updates, aligning composed queries with target captions while preserving source evidence through visual supervision. We further introduce Chord scoring, which combines target similarity with source-relative directional agreement in the frozen image space. Results across four ZS-CIR benchmarks show that combining target-caption supervision with source-image evidence leads to strong retrieval performance across datasets, backbone scales, and external galleries. The code is available on https://github.com/sehyunkwon/PACT.

Fri 25 SeptComputer Vision and Pattern Recognition
The gist
Finding images that match a user’s description of how to change a picture is hard, especially when we lack examples of the final images. The authors propose a new training method that uses not only text descriptions of the target images but also visual details from the original picture. This helps the system remember important parts of the source image while applying changes described in text. They also introduce a new way to score matches that balances following the text and keeping the original picture’s look. Their tests show better image search across different settings without needing new example images.
Open → 2609.31202v1

Privacy and stability improve slate recommendation with noisy scores

Decoupled Learning and Selection in Slate Recommendation for Privacy and Stability Under Noisy Scores

Abstract: We formalize slate recommendation as a randomized score learner followed by deterministic selection. First, an appropriately scoped differential-privacy guarantee passes through selection and its audit trace by post-processing. End-to-end privacy holds only when selector inputs are public or independent, previous private outputs, or separately privacy-accounted; fixing raw state or candidate information instead yields only a conditional guarantee. Second, we derive a logged margin certificate: bounded score-induced objective movement below half the smallest greedy decision margin guarantees that the ordered slate is unchanged. Controlled fixed-margin tests show near-linear exponent scaling, with an empirical slope of $-0.220$ (95% CI $[-0.231,-0.210]$) against the independent-noise reference $-1/4$. Real-anchor experiments on OULAD, MovieLens-25M, and Amazon Musical Instruments show that greater anchor weight reduces score-noise-induced ranking churn. OULAD and EdNet certificate checks validate the implementation of the logged inequality, while closed-loop simulations show bounded target drift and setting-dependent downstream utility. The contribution is therefore a privacy-scope contract and a certifiable score-to-slate stability mechanism, not a universal utility claim.

Thu 24 SeptMachine LearningInformation Retrieval
The gist
Slate recommendation means picking a list of items to show to users, like movies or products, based on scores given by a model. The authors explore how privacy rules apply when the system both learns those scores privately and then selects items based on them. They also develop a way to check if small random changes in scores will change the chosen list, helping guarantee stability. Their tests on real datasets show their method reduces ranking changes caused by noise without hurting performance too much.
Open → 2609.29453v1

Verifiable agentic environments improve long horizon reinforcement learning

Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms

Abstract: Language-model agents increasingly face long-horizon tasks with evolving state, interdependent decisions, and delayed outcomes. Scaling their training requires diverse agentic environments, dependable outcome signals, and low extension cost. Existing generation pipelines commonly construct an environment before defining its outcome rule or annotating its trajectories, leaving dynamics and evaluation to be aligned post hoc. VHD-Play reverses this dependency by sampling and solving a mathematical model before a corpus-grounded setter renders its decision process as stateful tools. The executable dynamics and trajectory-scoring reference are inherited from the same solved model. The pipeline produces 3,300 diverse agentic environments at a cost of a few cents each. Training Qwen3.6-35B-A3B on three families raises its mean agentic score from 0.204 to 0.815 in a five-family diagnostic. Gains also appear on held-out instances from all three training families and eight unseen mechanism families, then extend beyond the generated substrate to external benchmarks for general function calling, travel planning, and 365-day e-commerce. On E-Commerce Bench, the trained checkpoint completes every run without bankruptcy and exceeds Qwen3.7-Max. We compare written-out problems with stateful versions that reveal or hide their parameters. The comparison shows that most of the learnable gap lies in stateful interaction rather than underlying problem solving. A frozen 35B setter realizes larger environments, and scale-matched training retains gains as mechanism size and horizon grow, indicating the potential for an evolving training substrate.

Wed 23 SeptArtificial IntelligenceComputation and Language
The gist
Training language-model agents to handle long, complex tasks is hard because you need many varied environments and clear signals about success. The authors developed VHD-Play, which creates environments by first solving mathematical models, then turning them into interactive setups agents can use. This method produces thousands of diverse and verifiable environments cheaply and helps train agents that perform much better on a range of tasks, including e-commerce and travel planning challenges. The results show that learning to interact with dynamic states is more important than just solving static problems.
Open → 2609.27321v1

Semantic user profiling methods compared for streaming recommendations

When LLM-Based User Profiling Adds Value in Production Streaming Recommendation

Abstract: Personalized recommendation depends critically on how user representations are constructed from historical behavior. Two paradigms have emerged for constructing semantic user profiles in content-based recommendation. First, aggregate methods derive user representations as numerical aggregates of semantic item embeddings. Second, LLM-based methods generate natural-language summaries of user preferences and encode them through a text encoder. Each paradigm can be combined with temporal disentanglement of recent versus historical behavior. LLM-based profile generation is significantly more expensive than aggregate approaches, raising the question of when this additional cost is justified. We present a systematic comparison of four semantic user-profiling strategies, factorially crossed across representation type and temporal handling, evaluated on a real-world production dataset. The comparison reveals how these strategies differ across user behavior types, across both accuracy and beyond-accuracy dimensions of recommendation quality, and across the temporal-window setting that governs the disentanglement.

Wed 23 SeptInformation Retrieval
The gist
Online services often suggest content based on what users liked before, but creating user profiles from past actions varies in complexity and cost. The authors compare two main ways to build these profiles: one that averages numerical representations of items, and another that uses large language models (LLMs) to write natural descriptions of user interests. They tested these methods on a real streaming service dataset, looking at how well each works depending on the user’s behavior and the timing of actions. Their findings help decide when the expensive LLM-based profiling is worth using compared to simpler methods.
Open → 2609.27183v1

System improves trustworthiness of images in entity matching

Knowing When to Trust Images: Reliability-Aware Multi-modal Entity Alignment

Abstract: The visual modality, i.e., images, plays a key role in multi-modal entity alignment (MMEA). Existing approaches often directly fuse the image with other modalities to align different entities. Although simple, such strategies overlook the potential noise in the images and their semantic misalignment with corresponding entities, resulting in suboptimal fusion and degraded performance. Addressing this, we propose a novel Reliability-Aware framework for MMEA (RA-MMEA), which assesses visual reliability and adaptively improves unreliable visual representations for robust entity alignment. The core lies in two modules, including dependency-aware visual reliability prediction (DA-VRP) and stability-regularized visual embedding generation (SR-VEG). The former aims to estimate the reliability of an image by leveraging multi-modal dependency within the entity, while the latter focuses on producing alternative visual representation conditioned on semantics encoded in textual modalities for multi-modal fusion. Compared to current methods, RA-MMEA enables more reliable visual representations for modality fusion, thereby improving performance. In extensive experiments, RA-MMEA achieves state-of-the-art results, verifying the importance of reliable visual modality for entity alignment and the effectiveness of RA-MMEA. The code and results will be released.

Sun 20 SeptComputation and LanguageArtificial IntelligenceComputer Vision and Pattern Recognition
The gist
Matching data about the same thing from different sources is hard when images are not reliable or don’t fit well with text. The authors developed a method that checks how trustworthy images are and adjusts their representations to better match text descriptions. This approach helps combine images and words more effectively, making it easier to correctly identify the same entities across sources. Their tests showed that considering image reliability improves matching accuracy.
Open → 2609.23267v1

Livestream videos get smarter at matching products with show moments

Grounded Product Understanding in Livestream Videos

Abstract: E-commerce livestreams have emerged as an important channel for presenting products to online consumers, containing multiple products whose information is scattered in different moments. This poses significant challenges for downstream product understanding applications, such as product-centric livestream clipping, where models need to identify the product and its relevant segments for information gathering. However, existing benchmarks for general product understanding typically evaluate product retrieval and temporal localization in isolation, leaving the critical correspondence between product identity and temporal evidence largely unassessed. To address this limitation, we introduce GPUB, a large-scale benchmark comprising 3,000 livestream instances with quality-controlled multi-moment temporal annotations and a catalog of over 31K fashion products. GPUB supports three evaluation tasks: the main task Grounded Product Understanding (GPrU) requires jointly identifying the target product and localizing its supporting moments from a livestream video and a candidate product set; Product Retrieval and Product Moment Localization serve as two complementary subtasks. Evaluation of existing multimodal models shows that GPrU remains highly challenging, with the best-performing baseline achieving only 10.13% Pair mAP@.3. To narrow the performance gap, we further develop UniPro, a unified product understanding model that derives product-aligned and temporally structured representations from shared multimodal encoding, improving Pair mAP@.3 to 21.53% while achieving 37.23% Joint R@1@.3 on GPrU.

Thu 17 SeptComputer Vision and Pattern Recognition
The gist
Online shopping livestreams often show many products at different times, making it hard to tell which product relates to which video moment. The authors created GPUB, a big new dataset that connects fashion products to specific moments in livestreams to help computers learn this connection. They also built a new model called UniPro that does better at identifying products and the exact video parts that show them. Despite improvement, the task remains difficult for current technology.
Open → 2609.20508v1

Efficient engine speeds up linking data for multi-step reasoning

Efficiently Linking Unstructured Data for Multi-step Reasoning

Abstract: Modern LLMs and AI agents increasingly support data engineering workflows that integrate evidence from unstructured sources. Such pipelines typically do data retrieval, integration, and ranking before proceeding to more complex agentic reasoning or actions, e.g., for scientific discovery. The core retrieval problem in these workflows jointly executes multi-attribute filtering, multi-vector search, exact relational joins, and thresholded embedding-similarity joins. Given a planned query and monotone scoring function, our DASE query engine constructs and ranks candidate evidence tuples. It comprises (i) a multi-step reasoning query model over structured predicates, multiple vectors, and relational links; (ii) SemJI, a sparse materialized embedding-similarity join index for rare near-neighbor pairs; and (iii) a co-designed execution layer that combines predicate-aware ANN traversal, batched access, and threshold-based score aggregation. On scientific-discovery workloads, DASE retrieves candidate evidence for multi-step reasoning queries 6x to 46x faster than strong RDBMS, rerank, and vector-database baselines at comparable recall; and for tasks that require semantic-operator post-processing, DASE acts as a high-recall prefilter that makes downstream LLM evaluation both cheaper and more accurate -- e.g., on SemBench E-Commerce it improves BigQuery quality from 0.67 to 0.80 while cutting cost from $2.42 to $0.54.

Wed 16 SeptDatabasesArtificial Intelligence
The gist
Working with unstructured data like text and images is hard when you want to combine multiple pieces of information for complex reasoning. The authors developed a system called DASE that quickly finds and ranks relevant pieces of evidence from scattered data by mixing different search techniques. This helps AI tools make better, faster decisions when using lots of messy data, like in scientific research or e-commerce. Their experiments showed DASE is much faster than popular databases and improves the quality and cost of AI evaluations.
Open → 2609.19491v1

Region aware retrieval improves image and text matching in large models

RegRet: Enhancing Region-Level Retrieval in Large Multimodal Models

Abstract: Region-level retrieval aims to align user-specified image regions with relevant regions or textual descriptions, playing a crucial role in realworld applications such as e-commerce product search and RAG. Although recent Large Multimodal Models (LMMs) have made significant strides in multimodal retrieval, they primarily focus on global-level tasks and struggle to capture effective region-level representations. To bridge this gap, we present RegRet, an LMM-based Region-level Retrieval framework that enhances the regional representations without compromising overall global retrieval performance. At its core, RegRet integrates a Region-Aware Encoder to capture detailed regional features while balancing them with the global background context. To further enhance the fine-grained understanding and discriminability of representations, we design a multi-stage training pipeline that includes detailed localized captioning and regional contrastive learning tasks. In addition, considering the absence of region-level contrastive training data and the limited diversity of evaluation tasks in current benchmarks, we introduce the REGMB benchmark. It comprises 225k contrastive pairs, covering four multimodal retrieval tasks. Extensive experiments validate the effectiveness of our approach. RegRet outperforms strong baselines in the zero-shot setting. Further training with contrastive learning leads to an average improvement of more than 20\% on both REGMB and public benchmarks, while achieving comparable or better results on global-level retrieval tasks.

Tue 15 SeptComputer Vision and Pattern RecognitionArtificial IntelligenceInformation Retrieval
The gist
Finding specific parts of an image and matching them to related words or other image parts is important but hard. The authors designed RegRet, which helps large multimodal models look more closely at image regions without losing the overall picture. They also created a big new dataset to train and test this regional matching. Their experiments show RegRet does a better job than previous methods, especially when it comes to detailed image parts.
Open → 2609.16847v1

Agent system diagnoses and improves recommendation algorithms at scale

AURA: Agentic Diagnosis and Refinement for Production Recommender Systems at Scale

Abstract: How and why does a recommender system fail the users it serves? Oftentimes, practitioners are left to improve their algorithms based on a combination of feedback from stakeholder teams, domain expertise, and insights from data analyses. Yet the nuances of how and where recommendations perform well or poorly for end users are difficult to discern from aggregate quantitative metrics. Whereas these metrics provide a high-level and incomplete picture, further granularity into the quality of recommendations and their patterns requires reasoning with domain understanding and objectivity, at scale. We contemplate this complex conundrum and describe a method and implementation that uses the latest AI agentic advances to provide actionable diagnoses and improvements for production recommender systems. We present AURA (Agentic Understanding and Refinement of recommender Algorithms), an end-to-end agentic system that performs qualitative evaluation at scale and can then generate improvements to our algorithms at the code level. Specialized agents read production engagement logs, from thousands of sessions to millions, and surface patterns and examples of how the recommender fails real users. The next step uses those diagnoses and context about the recommender's own code, data, and training pipeline to propose and implement refinements grounded in that codebase. We report the system design, initial tests on production data from two large consumer platforms at a major media-streaming company, safeguards, operational learnings, and early results toward a self-improving recommender system. Finally, the diagnostic gap AURA closes is not specific to streaming. The architecture is built to transfer: every domain-specific element enters through the configuration layer that already ported it between our two platforms. We map it concretely to e-commerce and online-retail recommendation.

Tue 15 SeptInformation RetrievalArtificial IntelligenceMachine Learning
The gist
Recommender systems suggest things like movies or products but sometimes fail to give good suggestions. The authors created AURA, an AI tool that looks deeply at how and where these systems make mistakes by checking large amounts of real user data. AURA can then suggest and even change the recommendation algorithms to fix problems in the code. This approach was tested on big real platforms and can be adapted for other areas like online shopping.
Open → 2609.16625v1

Multi-agent ai creates editable product reviews from user images

From Visual Feedback to Textual Reviews: A Multi-Agent Vision-Language Framework for Image-Grounded Review Assistance

Abstract: Visual feedback in the form of user-uploaded images and videos is becoming increasingly common in e-commerce platforms because it provides authentic evidence of product quality, defects, packaging conditions, and real-world usage. However, visual feedback alone often lacks the contextual explanations and subjective opinions necessary for informed decision-making, while many users provide limited textual feedback due to the effort required to compose detailed reviews. To bridge this gap, we introduce image-grounded review assistance, a novel task that aims to generate editable review drafts from user-uploaded product images. Unlike conventional image captioning, which focuses on objective visual description, the proposed task requires product-specific understanding, sentiment estimation, and evidence-driven review composition under challenging real-world conditions, including degraded image quality, excessive zoom-in, target ambiguity, and partial product visibility. We propose a multi-agent vision-language framework consisting of four specialised roles: product grounding, visual sentiment estimation, visual evidence generation, and review synthesis. The framework employs explicit intermediate representations, including product entities, predicted ratings, and evidence summaries, to improve interpretability and visual grounding. Experiments on a curated subset of the Amazon Reviews Electronics dataset demonstrate the feasibility of generating coherent, product-aware, and sentiment-aware review drafts from visual feedback. To the best of our knowledge, this is the first study to formulate image-grounded review assistance as a multi-agent vision-language reasoning problem, providing a practical step toward AI-assisted review authoring in e-commerce systems.

Sun 13 SeptComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Shoppers often share pictures of products to show quality or defects, but don't always write detailed reviews. The authors introduce a system that looks at these photos and helps write draft reviews based on what it sees. Their approach breaks the task into steps, including recognizing parts of the product, guessing opinions, and putting together evidence for a review. Tests on real product photos show it can make sensible, opinion-aware reviews to help people write better feedback.
Open → 2609.14761v1

ShopEase system improves customer support with dense retrieval AI

ShopEase: A Generative AI-Based Multi-Agent Framework for Intelligent Enterprise Customer Support Using Hybrid Retrieval-Augmented Generation

Abstract: Enterprise customer support systems must answer customer questions correctly, retrieve the right policy information, use customer context, and pass difficult cases to human agents when needed. This paper presents ShopEase, a Generative AI-based multi-agent framework for enterprise customer support. The system combines six components: Intent, CRM, Memory, Hybrid RAG, Escalation, and Supervisor, and uses LLaMA 3.2 running locally through Ollama for response generation. The retrieval module combines FAISS (dense retrieval) and BM25 (sparse retrieval), and six configurations are evaluated: BM25-only, FAISS-only, Fair RRF, Weighted RRF, RRF with Cross-Encoder, and Top-10 Hybrid with Cross-Encoder. Instead of using a fixed mapping between intent and policy, the policy category is decided directly from the retrieved documents. The system was evaluated on 2632 held-out customer queries across six categories: Refund, Return, Shipping, Cancellation, Damaged Product, and Unknown. FAISS-only achieved the highest accuracy of 85.37\% (2247 correct predictions), closely followed by Weighted RRF at 85.07\%. BM25-only achieved only 55.74\% accuracy. Adding cross-encoder reranking did not improve results: RRF with Cross-Encoder reached 83.24\%, and Top-10 Hybrid with Cross-Encoder reached 81.88\%, while also increasing response latency. Category-level analysis shows strong performance on Shipping, Cancellation, and Return, while Unknown queries remain the main source of errors. Statistical testing using McNemar's test shows no significant difference between FAISS-only and Weighted RRF, though both perform significantly better than Fair RRF and the cross-encoder configurations. Overall, dense retrieval gives the best accuracy on this dataset, and additional reranking adds processing time without improving classification performance.

Sat 12 SeptComputation and Language
The gist
Enterprise customer support systems need to answer questions accurately and know when to ask a human for help. The authors created ShopEase, which uses multiple AI tools working together to handle customer questions better. Their system finds the right policies using a combination of dense and sparse search techniques and decides responses based on retrieved documents. Tests showed that dense search techniques worked best for understanding customer issues, and adding extra steps slowed things down without improving results.
Open → 2609.13856v1

Generative transformer improves relighting for 3D object images

RelightFormer: Feed-forward Generative Transformer for Multiview Object Relighting

Abstract: Image relighting is traditionally tackled via complex inverse rendering pipelines, which suffer from ill-posed optimization, or single-image generative models that ignore crucial multi-view cues necessary for understanding 3D geometry and material interactions. To address these limitations, we introduce a feed-forward generative Transformer for direct single- and multi-view image relighting that entirely bypasses explicit intrinsic property estimation. Adapted from a video foundation model, our architecture features a latent illumination module that dynamically injects target environment maps into spatial features via cross-attention. Furthermore, we employ permutation-invariant positional encodings to symmetrically process unordered multi-view inputs without sequential bias. To train this robust data-driven model, we construct the massive Laval Objaverse Dataset (LOD), comprising 90K objects and 39K unique illuminations. Extensive experiments demonstrate state-of-the-art visual quality, photorealistic relighting quality, and strong zero-shot generalization across single-view, multi-view, and novel-view relighting tasks.

Mon 7 SeptComputer Vision and Pattern RecognitionArtificial IntelligenceGraphics
The gist
Adjusting the lighting in photos of objects from different viewpoints is usually tricky and slow. The authors created a new AI model that changes lighting directly on images without needing to figure out complicated 3D details first. They trained this model using a huge dataset of thousands of objects and lighting setups. Their approach works well on images from one or many angles and even for unseen views, producing realistic relighting.
Open → 2609.07414v1