Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,312In brief24More (on-topic)3,333Not relevant13,753Not assessed87Older351Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| On-Policy Residual Learning for Diffusion Language Models This paper introduces an on-policy residual learning framework with an entropy-contraction objective to improve generation speed and quality in diffusion language models. Ranked #182New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 17 Oct, 06:18 | On-topic Introduces on-policy residual learning to substantially improve diffusion language model throughput and serving efficiency. | |
| SkillHorizon: Constructing Long-Horizon Agent Tasks through Skill Composition This paper introduces an evaluation benchmark that generates complex, long-horizon agent workflows by composing verifiable business skills. Ranked #182New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a benchmark and programmatic evaluator for measuring long-horizon agent task performance. | |
| Sensitivity or Criterion? Measuring and Moving When Language Models Act This paper introduces a diagnostic framework using signal detection theory to measure when language model agents decide to use tools. Ranked #182New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Analyzes when language model agents decide to take actions and use tools. | |
| Suggested Is Not Superseded: A Pre-Registered Evaluation of a System-One Model as a Memory-Retirement Gate The paper evaluates long-term memory gating models to test whether they distinguish established factual updates from mere suggestions. Ranked #182Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Assesses reliability and premature claim retirement risks in long-term assistant memory systems. | |
| IsoVRBench: Video Reward Modeling Without Human Feedback The paper presents a benchmark of synthetic paired comparisons to evaluate video reward models across specific visual and motion dimensions. Ranked #182New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Evaluates multimodal reward models on distinguishing video properties. | |
| Lifted State Policy Optimization: Internalizing Self-Correction through Transition Geometry The paper proposes a policy optimization method that treats reasoning transitions as learned coordinates in an energy space to internalize self-correction. Ranked #183Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Enables single-pass reasoning matching test-time search with lower latency and fewer tokens. | |
| DSABench: Benchmarking Source Discovery, Coordination, and Execution for Multi-Database Agents This paper introduces a benchmark to evaluate how language agents discover, coordinate, and query across multiple distinct databases to answer questions. Ranked #183New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a comprehensive benchmark evaluating discovery, coordination, and execution in database agents. | |
| Predicting how model propensities generalize out of distribution This paper studies how language model behavioral propensities generalize out-of-distribution across different elicitation methods. Ranked #183Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Investigates how alignment and elicitation can inadvertently generalize toward unsafe behavioral traits. | |
| Sequent-Prover: Training Agents for Formal, Checkable SMT-based Reasoning The paper introduces Sequent-Prover, training language model agents to perform multi-turn formal reasoning using an external SMT solver scratchpad. Ranked #183New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Trains language-model agents that reason, call external SMT solvers, and repair formalizations iteratively. | |
| VL-DocIR: A Benchmark for Vision-Based Long Document Information Retrieval Introduces a visual document retrieval benchmark focused on multi-page and cross-document evidence configurations in long documents. Ranked #183New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Evaluates visual document understanding and retrieval across multi-page rendered image documents. | |
| Target-Token Asymmetry Can Masquerade As Depth Specialization The paper shows that differences in predicted target tokens, such as whitespace, can falsely look like layer specialization across different corpora. Ranked #184Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Demonstrates methodological pitfalls in layer-ablation evaluations across corpora. | |
| Hindsight-Privileged RLVR: Single-Rollout Token Credit Without Group Sampling Derives token-level credit from single rollouts using hindsight future context rather than sampling multiple attempts per prompt. Ranked #184New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Improves RL training and sample efficiency without large sampling groups. | |
| Calibrating LLM Explanations to Human Decisions with Reinforcement Learning The authors introduce a reinforcement learning method that optimizes language model explanations specifically to improve human downstream decision accuracy. Ranked #184Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Optimizes model explanations to improve human decision accuracy and reliability in deployment. | |
| Can agents whose sole duty is to keep humans alive be hijacked to kill? Delayed multi-turn derived-action trap attacks in simulated high-stakes environments Tests delayed prompt injection attacks on life-support control agents in a simulated lunar habitat to see if safety duties persist. Ranked #184New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates failure modes and prompt injection vulnerabilities in tool-using autonomous AI agents. | |
| Answer It or Not? Towards Prompt-Adapted Adversarial Images on Vision Language Models Generates prompt-adapted adversarial images that block unauthorized visual questions on vision-language models while still answering legitimate prompts correctly. Ranked #184New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Addresses prompt-adapted adversarial images directly targeting vision-language models. | |
| SEA-HELM Audio: Evaluating Multilingual ASR-LLM Pipeline Performance Across SEA-HELM Subsets The paper presents an audio evaluation benchmark assessing accuracy drops between speech-input and text-input pipelines in Southeast Asian languages. Ranked #185New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Introduces an evaluation benchmark measuring performance degradation in multilingual ASR-LLM pipelines. | |
| Divergent Paths: Learning an Outcome-Predictive Representation of LLM Agent Trajectories The paper learns trajectory embeddings to predict the success or failure of language model agents early during execution. Ranked #185New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Cuts training FLOPs by up to 47.3% via early stopping in agent policy optimization. | |
| LLM Evaluation is a Forecasting Problem This paper frames language model evaluation as a forecasting problem, using time-series techniques on historical data to predict post-deployment performance. Ranked #185New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Studies temporal performance drift of LLMs to improve reliable model selection in deployment. | |
| Structural Chain-of-Thought: Reasoning with Executable Computational Graphs Replaces free-form reasoning text with executable computational graphs trained via reinforcement learning to improve reasoning speed and verifiability. Ranked #185New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Enables models to plan and reason using executable computational graphs. | |
| VisExMEM: Visual Evidence-Based Explainable Multimodal Evaluation Metric Presents a reference-free evaluation metric that decomposes captions into atomic claims and verifies them against visual evidence. Ranked #185New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Evaluates vision-language models and image-text alignment using visual evidence and claim decomposition. | |
| Lost in Translation: Why Lower Bias Scores Don't Mean Fairer Models This paper evaluates multilingual bias audits across ten language models, showing that lower stereotype scores often reflect translation artifacts rather than fairer behavior. Ranked #186Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Critiques multilingual bias benchmarks, revealing why lower stereotype scores reflect translation noise. | |
| Frustratingly Easy Majority Voting Using Length Conditioning The paper proposes weighting majority voting responses using a length-conditioned prior adapted at test time without model fine-tuning. Ranked #186Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Improves test-time inference efficiency and accuracy of reasoning without model fine-tuning. | |
| On the Calibration of Crowdsourcing Aggregators The paper provides theoretical and empirical analyses of calibration in crowdsourced label aggregation across human and language model datasets. Ranked #186Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Examines miscalibration and overconfidence in probabilistic aggregator models. | |
| SkillHorizon: Constructing Long-Horizon Agent Tasks through Skill Composition This paper introduces an evaluation benchmark that generates complex, long-horizon agent workflows by composing verifiable business skills. Ranked #186New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates multi-step LLM agents across long-horizon business tasks requiring shared state updates. | |
| Look Back to Move Forward: Learning to Revisit for Agentic Compositional Image Generation Develops an agentic text-to-image framework allowing agents to revisit and select earlier image states for iterative compositional generation. Ranked #186New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Focuses on text-to-image models and iterative multimodal visual reasoning. | |
| OpenTWBench (Open Taiwan Benchmark): Taiwan-Domain Knowledge Evaluation for Large Language Models, and How the Measurement Apparatus Shapes the Score The paper introduces a two-part benchmark to evaluate language models on general and examination knowledge concerning Taiwan. Ranked #187Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Introduces an open evaluation suite and benchmark datasets for Taiwan-domain knowledge in LLMs. | |
| Does the Linguistic Register of Thinking Matter? Wenyan as a Compact Reasoning Register for Chinese LLMs This paper investigates using Classical Chinese as a compact intermediate reasoning register to reduce token length in Chinese language models. Ranked #187Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Compresses reasoning token length to reduce inference cost and decoding budget. | |
| PVFH-Bench: Benchmarking Premise-Induced Visual Fixation Hallucinations in Vision-Language Models This paper introduces a benchmark to measure vision-language model hallucinations caused by deceptive textual premises paired with visually confusable images. Ranked #187Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Addresses reliability and object hallucination risks in deployed vision-language systems. | |
| A Reference Architecture for Integrating Agentic AI into Enterprise Software: Responsibilities, Interactions, and Variability This paper presents a technology-neutral reference architecture for integrating agentic AI capabilities into enterprise software systems. Ranked #187New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Presents a comprehensive reference architecture for integrating multi-action agentic AI into enterprise software. | |
| ELECT: Settling Identity Before Rendering for Scalable Story Visualization The paper proposes a training-free method establishing entity identity before final rendering to maintain visual consistency across long storyboards. Ranked #187New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Generates coherent multi-frame visual scenes conditioned on text stories. | |
| Visual Sanad: Auditing Vision-Language Models for False Acceptance of Qur'anic Citations The paper introduces a benchmark to measure how often vision-language models accept incorrect Qur'anic citations across various visual presentations. Ranked #188New testPreprint6/10 suggested | Multimodal models via arxiv · 17 Oct, 06:18 | On-topic Audits vision-language models on citation verification under varied visual framings. | |
| MemCut: Every Secret Leaves a Trail in the Long-Term Memory of LLM Agents The paper introduces a benchmark to evaluate whether language model agents reliably delete sensitive user data and secret evidence from memory upon request. Ranked #188New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17 | On-topic Introduces MemCut, a benchmark and metrics evaluating whether LLMs forget sensitive user data. | |
| Scaling LLM Inference via Parallelism: Throughput-Latency Trade-off The paper presents a systematic framework for selecting optimal parallelism strategies to balance latency and throughput during distributed language model serving. Ranked #188New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Provides a systematic framework for efficient distributed LLM serving and throughput-latency optimization. | |
| BrowseSafe++: Benchmarking Prompt Injection Detection Across Web Contexts, Languages, and Browser Views This paper introduces a benchmark of paired web pages to evaluate prompt injection detection across browser agents and views. Ranked #188New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Focuses on detecting prompt injections and securing browser agents against web attacks. | |
| Iterate on an LLM, Serve on an SLM The paper proposes prototyping agent workflows with large language models and distilling converged live production interactions into smaller models. Ranked #188New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Demonstrates prompt iteration and production distillation for task-specific agent workflows. | |
| More Than Seeing: Vision as Scalable Anchors in Native Multimodal Pre-Training The paper studies how visual input signals affect scaling behavior and non-visual reasoning during native multimodal pretraining. Ranked #189Surprising resultPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Investigates native multimodal pre-training and how visual inputs affect reasoning. | |
| Titan-Llama: Retrofitting Pre-trained LLMs with External Neural Memory Modules The paper presents a method to retrofit attention models with hierarchical test-time neural memory modules to improve long-context inference efficiency. Ranked #189New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Introduces fixed-size neural memory modules to reduce long-context KV cache inference overhead. | |
| From Unreliable to Usable: Decomposing Failure Modes and Confidence Signals in LLM Entity Extraction - Evidence from Cyber Threat Intelligence The paper evaluates failure modes and confidence calibration across frontier and lightweight language models during entity extraction from cyber threat intelligence reports. Ranked #189Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Conducts a frozen audit of LLM extraction and confidence calibration. | |
| Poisoned into Action: Context Attacks on Typed Decision Workflows The paper studies how context poisoning triggers unauthorized state changes in typed decision workflows and evaluates an execution-aware defense protocol. Ranked #189New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Evaluates context poisoning attacks and mitigation gates that cause models to trigger policy-violating automated actions. | |
| A Self-Defining Operator for Distributed Applications The paper presents an autonomous agent architecture that distills distributed system incident resolutions into reusable detectors and operational playbooks. Ranked #189New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Presents an LLM agent that investigates incidents, uses tools, and updates operational playbooks. | |
| STP-GUI: Spatially-Aware Token Pruning for Efficient Web-GUI Agents This paper introduces STP-GUI, a spatially aware token-pruning framework that preserves local interface structures for efficient web-GUI vision-language agents. Ranked #190New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Optimizes token efficiency for vision-language models interacting with web interfaces. | |
| Cascaded Retrieval with Cross-Modal Query Refinement for Efficient Visual Document Search This paper proposes SCOUT, a cascaded two-track framework combining visual candidate filtering, offline semantic summaries, and query refinement for document retrieval. Ranked #190New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Achieves up to 80x query latency speedups through cascaded retrieval and candidate filtering. | |
| Are AI Decision Models Biased? A Case Study of Jev The paper evaluates an artificial intelligence decision model across cognitive reflection, classical bias benchmarks, and pricing tasks. Ranked #190Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Audits cognitive, social, and statistical biases in AI decision-making models using behavioral benchmarks. | |
| ConformalGuard: False-Positive-Controlled Action Gating for LM Classifiers The paper proposes a conformal calibration layer to provide finite-sample control over false positives in language model classification workflows. Ranked #190New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Proposes conformal gating to control false positive rates in high-stakes operational LM deployments. | |
| DSABench: Benchmarking Source Discovery, Coordination, and Execution for Multi-Database Agents This paper introduces a benchmark to evaluate how language agents discover, coordinate, and query across multiple distinct databases to answer questions. Ranked #190New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates autonomous agents coordinating, discovering sources, and executing across multiple databases. | |
| Progress-Aware Distillation for Mitigating Stagnation in Small Language Model Agents The paper introduces a progress-aware distillation method to prevent small language model agents from stalling in repetitive loops during long tasks. Ranked #191New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 15:06 | On-topic Addresses failure modes and looping stagnation in deploying autonomous language model agents. | |
| What Breaks in Multi-Image Schema Application? Factorial Oracle Interventions Reveal an Instance-Information Bottleneck Uses factorial interventions to locate failure points in multi-image vision-language reasoning pipelines. Ranked #191Surprising resultPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Examines failure modes and instance retrieval bottlenecks in multi-image vision-language models. | |
| Scaling Multilingual Mixture-of-Experts: How Language Coverage Changes Capacity and Compute Allocation This paper proposes a joint scaling law evaluating how adding experts in multilingual mixture-of-experts models affects language capacity and loss penalties. Ranked #191New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Analyzes compute allocation and capacity scaling in mixture-of-experts architectures. | |
| Plan or Retrieve First? A Scale Study of Tool Discovery Architectures Compares retrieve-then-plan and plan-then-retrieve strategies for tool discovery across varying tool registry complexities. Ranked #191Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Investigates tool discovery and execution planning architectures for tool-augmented language model agents. | |
| CoRedSkill: Stealthy Skill Composition Attacks via AST-Guided Injection The paper presents a red-teaming method that injects stealthy attack operations across interdependent agent skills using code rewriting. Ranked #191New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Presents a benchmark evaluating agentic skill composition attack success. | |
| MemCut: Every Secret Leaves a Trail in the Long-Term Memory of LLM Agents The paper introduces a benchmark to evaluate whether language model agents reliably delete sensitive user data and secret evidence from memory upon request. Ranked #192New testPreprint7/10 suggested | AI agents via arxiv · 17 Oct, 06:17 | On-topic Evaluates long-term memory management and user-requested forgetting in LLM conversational agents. | |
| TRACER: Trace-Regularized Training Against Indirect Prompt Injection The paper introduces an adversarial training framework that hardens models against indirect prompt injection across retrieval and tool interaction channels. Ranked #192Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 15:06 | On-topic Proposes training techniques to defend language models against indirect prompt injection attacks. | |
| CORE: Contrast-and-Reflect for Self-Supervised Prompt Optimization in Anomaly Synthesis Optimizes multimodal prompts self-supervisedly to synthesize realistic industrial defect images for anomaly detection. Ranked #192New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Employs large multimodal models via self-supervised prompt reflection for controllable image generation. | |
| Delta Tokenizer: Reference-Identical Tokenization of Session Deltas and Segment-Parallelism for Long Prompts on Multicore CPUs The paper introduces a CPU front-end tokenizer that encodes only session deltas to speed up multi-turn inference. Ranked #192New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Introduces an efficient CPU tokenization system to accelerate serving and reduce time to first token. | |
| How GUI Agents Fail: GraceBench for Execution Stability Diagnosis The paper introduces an evaluation framework to diagnose step-level execution stability and error recovery in graphical user interface agents. Ranked #192New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Introduces GraceBench to benchmark step-level trajectory stability and failure modes in GUI agents. | |
| The Gap Is in the Geometry: A Convergent Meta-Analysis of Multi-Facet Aerial and Remote Sensing Video Understanding A meta-analysis shows multimodal language models struggle with geometric acquisition variations in aerial video, exposing limitations in standard QA-based evaluations. Ranked #193Surprising resultPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Analyzes multimodal large language models on video understanding tasks across various spatial conditions. | |
| RouteWeaver: Weaving Mode Selection and Execution into Unified LLM Routing The paper introduces a unified routing framework that learns both execution mode selection and model assignment across diverse tasks. Ranked #193New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Optimizes LLM query routing across heterogeneous models to balance capability and serving efficiency. | |
| Typed Decision Models on Tabular Data The paper compares typed probabilistic decision models with large language models on tabular data classification and format robustness. Ranked #193Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Evaluates LLM and decision model robustness under 21 input rewrite variations on tabular data. | |
| Drivers of LLM Consensus Formation Under Competing Incentives: A Simulation Study The paper evaluates how open and closed language models reach consensus when presented with competing payoff incentives. Ranked #193Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Studies how networks of LLM agents coordinate and reach consensus under competing incentives. | |
| Fairness under Prediction Error in Predict-then-Optimize Pipelines The paper analyzes how prediction errors affect demographic parity fairness violations in budget-constrained allocation problems. Ranked #193Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Analyzes fairness constraint violations and reliability under prediction errors in deployed decision pipelines. |
1,261–1,320 of 17,312← PreviousNext →