Papers for

content management teams

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.

Collection date metadata strongly affects temporal question answers

Collection-Date Assignment Redirects Answers in Temporal QA

Abstract: A collection date records when a source revision was acquired, not when its contents changed. Can assigning this metadata redirect temporal question answering despite unchanged evidence? Swapping the question day and following day between two revisions lowers later-reference exact match (EM) by 37.6-51.4 percentage points across five open-weight LLMs on 595 Wikipedia pages, replicated on 128 disjoint pages. Previous-day swaps retain 29.0-39.6-point losses on 497 pages. Seven prompt variants, including explicit ignore-collection instructions and archived-revision provenance, retain the effect. Fictional answers and English month-name question dates show that remembered answers and identical date strings are unnecessary. Matched thinking reduces Qwen3.5's penalty from 30.1 to 11.1 points, while native reasoners vary. Two API LLMs show small Wikipedia swap effects and rarely shift to earlier answers on fictional records. Large swap-induced losses concentrate around the question date; swaps between distant dates have much smaller effects. Adding fields containing one observed acquisition date lowers EM relative to omission in three of five LLMs after correction. The diagnostic isolates metadata effects from changing evidence or missing the relevant revision, and provides a practical audit for temporal QA inputs.

Fri 23 OctComputation and Language
The gist
When computers answer questions about when facts happened, they often look at the date when they got the information, not when the fact itself changed. The authors show that changing the assigned collection date (the date the information was taken) can make language models give very different answers, even if the facts haven’t changed at all. This means these models might rely too much on metadata dates rather than actual content changes. The authors tested several large language models and found consistent effects, offering a method to check how dates influence answers about time.
Open → 2610.32191v1

Vision language models improve reranking with efficient decision making

Multimodal Reranking Is a System One Decision

Abstract: Multimodal rerankers are usually dedicated models: a vision-language model (VLM) is trained on retrieval data to score the candidates of a query. Scoring options is not specific to reranking. System One decision models such as Jev map a state and a typed question to one probability per option without generating text, and general models can make such decisions with little or no training. Reranking is one such decision, with the query and its candidates as the state and the candidates as the options. We therefore argue that a general VLM can serve as a multimodal reranker through System One decisions, and that decision tuning can improve its reranking. We ask a VLM one choice question over the top 16 candidates of a first-stage retriever and rank the candidates by the probabilities of their labels, read from one forward pass. On MMEB-V2 image retrieval, this choice scores 5.4 to 7.3 hit@1 above a yes/no question per candidate, and the yes/no question falls below the first stage. With the choice, zero-shot Qwen3.8-27B beats the supervised Qwen3-VL-Reranker-2B on all three benchmarks and comes within 1.6 points of Reranker-8B on each. It matches Reranker-8B on MMEB-V2 visual documents (84.79 vs. 84.78 nDCG@5), at 1.1 to 1.5 times its latency. JEV-27B-VL, the same backbone tuned on text decisions only, improves page reranking by 0.6 to 1.1 points, with gains on most tasks (sign test, p <= 0.017), and leaves natural images unchanged. The zero-shot model is larger than the supervised rerankers and of a newer generation, so these comparisons are confounded; we discuss this and other alternative views, and the experiments that would test the position.

Tue 20 OctInformation Retrieval
The gist
Finding the best answer among many usually requires special models trained to do just that. This paper shows that general vision-language models (VLMs), which understand images and text, can also pick the best choice efficiently by framing it as a simple decision problem. The authors found that just asking one clear question about the top options lets these general models rank choices better than some trained models, even without special retraining. They tested this on several image retrieval tasks and found it matches or beats more specialized methods, suggesting a new way to use big vision-language models for reranking.
Open → 2610.28435v1

Optimizing document endings to guide AI answers on facts

Shaping the Document, Not the Model: Optimized Suffixes for Context Following

Abstract: When a user supplies a document and wants the answer drawn from it, a language model is expected to follow it even when it contradicts its parametric knowledge. Prior work has focused on making the model rely on context by steering activations, contrasting logits during decoding, or instructing it in plain language. We ask whether the document alone, with no instruction written into it and no access to the model at inference time, is enough. We optimize a document suffix by discrete token search, starting from the answer to a single question and extending the objective to a set of passages, to multi-hop questions that follow from the planted fact, and to later turns of a conversation. On counterfactual passages that contradict parametric knowledge of the model, the suffix flips nearly all items the model resists across three instruction-tuned model families, in most settings outperforms the decoding baseline and the strongest instruction we could write, and matches activation steering. Controls show that the suffix largely acts through the passage rather than injecting the answer, and a single suffix per model transfers to unseen passages. Each extended objective carries the effect to its target, so the objective, not only the document, sets how far the effect reaches.

Mon 19 OctComputation and Language
The gist
Sometimes AI language models ignore the information given in a document and rely on their own internal knowledge, even if it's wrong. The authors found a way to add special endings to documents that help the AI follow the document's facts instead. They do this by searching for the best sequence of words to add after the document, without changing the AI itself or giving extra instructions. This technique helped AI models better answer questions based on the document, even when the document contradicts what the AI originally 'knew.'
Open → 2610.26236v1

Machine learning finds layered meanings in graphs with texts

Uncovering Semantic Hierarchies in Text-Attributed Graphs via Variational EM-based LLM-GNN Synergy

Abstract: While hierarchical structures are intrinsic to organizing human knowledge, current representation learning on text-attributed graphs predominantly operates on flat semantic spaces, overlooking the rich coarse-to-fine granularity inherent in real-world data. To bridge this gap, we propose SHiFT, which synergizes the reasoning power of LLMs with the structural representation capability of GNNs within an Expectation-Maximization (EM) paradigm. In the E-step, we harness LLMs as expert taxonomists to induce an explicit semantic tree, introducing a lightweight Mapping & Diagnosis strategy that reduces inference costs by surgically updating the hierarchy only when anomalies are detected. In the M-step, we align the GNN encoder with this induced hierarchy from complementary perspectives, including clustering partition, topological skeleton, and semantic concepts. Theoretical analysis interprets SHiFT as a generalized Variational EM algorithm. Extensive experiments demonstrate that SHiFT not only achieves state-of-the-art performance but also uncovers high-quality, human-readable taxonomies. Our code can be found at https://github.com/Cloudy1225/SHiFT.

Sat 17 OctMachine Learning
The gist
Many real-world ideas are organized in layers, from broad categories to specific details, but most existing methods ignore these layers when analyzing text connected in networks. The authors propose a method called SHiFT that combines large language models (LLMs) and graph neural networks (GNNs) to discover and organize these layers automatically. SHiFT uses a two-step approach where LLMs first build a semantic tree of concepts and then GNNs learn to represent these concepts aligned with the tree structure. This approach uncovers clear and understandable hierarchies in data and performs well in experiments.
Open → 2610.24624v1

Large language models expand analysis and research workflows in cultural studies

Large Language Models, Machine-Assisted Cultural Analytics, and the Agentic Turn

Abstract: Technologies like large language models make it easier to convert cultural material into analyzable variables, while LLM-driven agentic systems extend machine assistance to other parts of the research workflow. Rather than a full how-to guide or an exhaustive review of recent LLM use cases, this chapter looks at some crucial aspects of this new paradigm as well as the bigger picture. I argue that these technologies expand the scale, methodological range, and scope for delegation, while the basic logic by which those measurements support cultural claims remains mostly unchanged.

Thu 8 OctComputers and SocietyComputation and Language
The gist
Studying culture involves turning stories, art, and history into data we can analyze. The authors explain how big AI language models help to do this faster and across more types of data without changing how we draw conclusions from that data. They also show how AI can take on more tasks in research workflows, making studying culture easier and broader. This is not a detailed guide but a discussion about what this new AI use means overall.
Open → 2610.12792v1

Model learns to handle long texts with small context windows

Trained Agentic Context Management

Abstract: We study long context language models. Instead of training long context natively, or designing a long context harness, we train a model over the simplest possible harness: a tool to call itself with any specified prompt and a tool to read tokens in a range from the input context. We finetune Qwen3.6-35B-A3B on a diverse synthetic dataset using this harness. With only 8,000 tokens of context, our small model is as strong as GPT-5.4 with 1M tokens of context on the OOLONG-synth benchmark when document length exceeds 40K tokens.

Thu 1 OctMachine LearningComputation and Language
The gist
Reading very long texts is hard for language models because they need to remember lots of information at once. The authors found a way to train a model that can manage long documents by teaching it to call itself with smaller pieces of the text and read parts as needed. They trained a model this way on fake but varied text, and it became as good as a much bigger model that uses huge amounts of text at once. This means smaller models can work with very long texts effectively.
Open → 2610.02404v1

Knowledge pull requests improve continual document updates across languages

Knowledge Pull Requests for Continual Document Authoring

Abstract: We introduce Knowledge Pull Requests (KPRs), a framework for continual document authoring that makes each change interpretable. Documents require ongoing revision as new knowledge surfaces from other sources, languages, or times, but existing approaches either edit with no account of what knowledge changed or regenerate from scratch. A KPR integrates new knowledge into a document by extracting claims, filtering and routing them to sections, and flagging conflicts with existing content, producing a ChangeLog that separates what knowledge changes (claim proposal) from how the text changes (document diff). We evaluate KPRs on revising Wikipedia across languages and updating query-driven reports on RAGTIME. KPRs integrate more information and better preserve existing content than rewriting from sources or regenerating from scratch, while adding the most information per token generated. A KPR-revised article also grounds question answering better than a frontier model with search, which does not surface knowledge documented only in other languages.

Tue 22 SeptComputation and Language
The gist
Documents like Wikipedia articles need constant updates as new facts appear. The authors present a method called Knowledge Pull Requests (KPRs) that helps update documents by clearly showing what knowledge changes and how the text is altered. KPRs work by extracting new claims, deciding where they fit in the document, and marking contradictions with old content. This approach preserves more existing information and adds more useful facts than rewriting entire texts from scratch. It also helps answer questions better by including knowledge from documents in different languages.
Open → 2609.26634v1

Closed loop system greatly improves meeting text output targets

A Closed-Loop Control Architecture for Reliable Constraint Satisfaction in LLM Text Generation

Abstract: Software systems increasingly embed a large language model in features that must satisfy a numeric output constraint, that is, a requirement expressible as a number or an interval and checkable by code, such as a target word count or a target readability grade band. Because such a model is non-deterministic, is configured through natural-language instructions rather than a typed interface, and satisfies a stated requirement only approximately, a single prompt neither reliably meets the target nor preserves the source content. This paper presents and evaluates a closed-loop control architecture for this problem. It has five stages: generate, evaluate, adjust, archive, and analyze. The model is called only to write and to edit text, while deterministic code compares a composite readability value against a target band, rejects any edit that drops source entities, numbers, or keywords, and makes every accept decision. Over 114 single-shot generation jobs and 240 closed-loop runs on four commercial models, single-shot prompting met the target in 21.1 to 31.6 percent of cases and the closed loop in 92.5 to 98.8 percent, within two edit rounds on average and at a recall-based fidelity of 0.92 to 0.93; the two models common to both settings show the same effect. Because the controller optimizes the value on which success is scored, the result establishes reproducible control over a declared, computable metric and not validated human difficulty. The transferable practice is to declare the acceptance condition as code, bound the model to local edits, and gate every edit on a content check.

Thu 17 SeptSoftware Engineering
The gist
When using large language models to create text that meets specific numeric rules—like word counts or readability scores—one attempt often doesn’t get it right. The authors show a system that generates text, checks if it fits the rules, then edits it repeatedly until it does. This looping method made the output meet targets much more reliably and kept important details intact. The system works by using code to decide if edits are good enough, rather than relying on the language model’s guesswork alone.
Open → 2609.19710v1

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

Text summarization improves by focusing on important tokens

TIAO: Token Importance-Aware Policy Optimization for Text Summarization

Abstract: Text summarization requires models to condense content while preserving key qualities such as consistency and coherence. Large language models (LLMs) have shown strong performance on this task and can be further improved through reinforcement learning (RL). However, most existing methods apply reward signals directly to undifferentiated token sequences, overlooking the varying importance of individual tokens to word and sentence level quality in summarization. In this paper, we propose Token Importance-Aware Policy Optimization (TIAO), a novel reinforcement learning strategy that explicitly leverages token-importance awareness. Specifically, TIAO identifies core tokens based on token dependency and reweights a trajectory's advantage according to its overall dependencies. Experiments on the real world dataset show that our TIAO achieves highly competitive results, and that a 7B foundation model enhanced by TIAO performs comparably to GPT-4 and GPT-5-nano. Code is available at https://github.com/TechCloud-x/TIAO

Tue 15 SeptComputation and Language
The gist
Summarizing text means making it shorter but still keeping the important ideas clear and connected. The paper introduces a new method called TIAO that helps computer models recognize which words in a summary are most important. By paying special attention to these key words during training, the method improves how well summaries make sense and capture the main points. Experiments show that even smaller trained models using this method can perform as well as much larger AI systems like GPT-4. This approach refines how AI learns to summarize by understanding the role of each token in the sentence.
Open → 2609.16748v1

CausalChapter improves chaptering for long instructional videos

CausalChapter: Improving Long-Video Chaptering with Interventional Dependency Modeling

Abstract: Long-form instructional videos require automatic chaptering to support browsing, navigation, and knowledge access. Recent long-context language models can perform chaptering from textualized video inputs, but they remain costly and brittle for content-dense lecture videos with long transcripts, smooth topic transitions, and detailed chapter outputs. A scalable segment-then-caption paradigm reduces this cost, but introduces two new challenges: boundary error propagation and fragmented cross-chapter context. We propose \textbf{CausalChapter}, an intervention-inspired framework for long-video chaptering that estimates prediction-level influence through lightweight masking and removal interventions. For boundary localization, our Local Dependency Shift module detects drops in predictive dependency between adjacent temporal windows; for chapter description generation, our Cross-Segment Support Selection module reranks historical contexts according to their support for the current prediction. Experiments on long-video chaptering benchmarks show that CausalChapter improves boundary localization, chapter description quality, and cross-chapter coherence.

Tue 8 SeptComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Long instructional videos can be hard to navigate because they often don’t have clear chapters to separate topics. The authors propose CausalChapter, a method that helps find chapter boundaries and write chapter summaries more accurately by checking how different parts of the video influence predictions. This approach uses a clever masking technique to spot where topics change and improves linking across chapters for better descriptions. Their method works better than previous ones for videos with long, dense content and smoothly changing topics.
Open → 2609.08686v1