Every paper checked — nothing hidden

All papers

Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.

All17,313 In brief24 More (on-topic)3,333 Not relevant13,754 Not assessed87 Older352 Could not confirm93 Blogs with no paper266 All blogs284

Paper	Topic	Relevance
Matoi: Selective Revalidation of Consequence Evidence for Agent Authorization The paper introduces an authorization system that reuses validated consequence evidence across agent execution branches while maintaining separate live authority. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on authorization architecture, not LLM benchmark evaluation.
When do quantum views help? Complementary classical-quantum representations in multi-view fusion The paper studies combining classical and quantum representations to determine when multi-view fusion improves machine learning predictions over single representations. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Investigates quantum ML representation learning, not AI safety, jailbreaks, or guardrails.
When do quantum views help? Complementary classical-quantum representations in multi-view fusion The paper studies combining classical and quantum representations to determine when multi-view fusion improves machine learning predictions over single representations. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on classical-quantum machine learning fusion, not LLM agents or tool use.
When do quantum views help? Complementary classical-quantum representations in multi-view fusion The paper studies combining classical and quantum representations to determine when multi-view fusion improves machine learning predictions over single representations. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Studies quantum feature fusion algorithms, not LLM benchmarks or evaluation methods.
Watts If? A Benchmark for Counterfactual Reasoning About the AI Data Center Buildout The paper introduces a benchmark to evaluate how language models perform counterfactual reasoning regarding AI data center infrastructure and siting commitments. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Evaluates domain-specific counterfactual reasoning rather than AI deployment safety or guardrails.
Watts If? A Benchmark for Counterfactual Reasoning About the AI Data Center Buildout The paper introduces a benchmark to evaluate how language models perform counterfactual reasoning regarding AI data center infrastructure and siting commitments. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Introduces an infrastructure decision benchmark, not autonomous agent planning or tool use.
Multi-Reward Hacking: From Aggregation-Induced Drift to Trust-Anchored Defense The paper analyzes multi-reward reinforcement learning failures where auxiliary objectives degrade primary performance and introduces a trust-anchored gating method. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on training algorithms rather than empirical benchmark development or LLM-as-a-judge.
Multi-Reward Hacking: From Aggregation-Induced Drift to Trust-Anchored Defense The paper analyzes multi-reward reinforcement learning failures where auxiliary objectives degrade primary performance and introduces a trust-anchored gating method. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Studies reinforcement learning post-training dynamics, not LLM multi-step agent tool use.
Multi-Reward Hacking: From Aggregation-Induced Drift to Trust-Anchored Defense The paper analyzes multi-reward reinforcement learning failures where auxiliary objectives degrade primary performance and introduces a trust-anchored gating method. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Addresses RL post-training reward alignment rather than model compression or inference efficiency.
Jev-LiteMAS: Separating Decisions from Generation in Multi-Agent Workflows The paper proposes a framework that replaces generative model calls with lightweight decision models for routing and reviewing in multi-agent workflows. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on multi-agent workflow architecture rather than evaluating general LLM benchmarks.
Jev-LiteMAS: Separating Decisions from Generation in Multi-Agent Workflows The paper proposes a framework that replaces generative model calls with lightweight decision models for routing and reviewing in multi-agent workflows. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Targets coordination efficiency in agent workflows rather than AI safety or hallucination defenses.
JevAgentHarness: Benchmarking, Deploying and Training Decision Models That Share Control with LLMs The paper presents a benchmark and harness to delegate routine tool choices from large language models to cheaper decision models. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on agent harness execution and cost rather than safety guardrails or jailbreaks.
Where, How, and When: Query-Guided Grounding for Video Understanding Presents methods for query-guided video grounding across object localization, relationship detection, and temporal moment identification. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Studies generic video action grounding, not industrial quality control or defect detection.
Where, How, and When: Query-Guided Grounding for Video Understanding Presents methods for query-guided video grounding across object localization, relationship detection, and temporal moment identification. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Studies video visual grounding algorithms, not AI safety, alignment, or guardrail systems.
Where, How, and When: Query-Guided Grounding for Video Understanding Presents methods for query-guided video grounding across object localization, relationship detection, and temporal moment identification. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Presents visual grounding architectures, not LLM benchmark design or evaluation analysis.
Where, How, and When: Query-Guided Grounding for Video Understanding Presents methods for query-guided video grounding across object localization, relationship detection, and temporal moment identification. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Deals with video visual grounding, not LLM agent planning or tool execution.
Do Agents Need a New Kind of Paper? SPRA: Structured Paper Representation for Agents Creates a structured schema that converts research papers into connected scientific entities to improve retrieval-augmented generation for autonomous agents. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Processes structured scientific text and relations, not multimodal vision, audio, or video data.
Do Agents Need a New Kind of Paper? SPRA: Structured Paper Representation for Agents Creates a structured schema that converts research papers into connected scientific entities to improve retrieval-augmented generation for autonomous agents. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Focuses on scientific literature representations, not software engineering, code synthesis, or repair.
Do Agents Need a New Kind of Paper? SPRA: Structured Paper Representation for Agents Creates a structured schema that converts research papers into connected scientific entities to improve retrieval-augmented generation for autonomous agents. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Proposes an information representation framework rather than general LLM evaluation benchmarks.
Do Agents Need a New Kind of Paper? SPRA: Structured Paper Representation for Agents Creates a structured schema that converts research papers into connected scientific entities to improve retrieval-augmented generation for autonomous agents. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Evaluates context token reductions, but does not address model compression, pruning, or quantization.
Do Agents Need a New Kind of Paper? SPRA: Structured Paper Representation for Agents Creates a structured schema that converts research papers into connected scientific entities to improve retrieval-augmented generation for autonomous agents. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on knowledge structuring and retrieval rather than AI safety guardrails or jailbreaks.
Learning from Interaction Experience via Localized Context Distillation Introduces a distillation method that identifies critical sub-spans in student agent trajectories for teacher revision and continual learning. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on post-training distillation methodology rather than LLM benchmarks or evaluation.
Learning from Interaction Experience via Localized Context Distillation Introduces a distillation method that identifies critical sub-spans in student agent trajectories for teacher revision and continual learning. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Targets general model post-training distillation, not safety guardrails or hallucination mitigation.
Learning from Interaction Experience via Localized Context Distillation Introduces a distillation method that identifies critical sub-spans in student agent trajectories for teacher revision and continual learning. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Introduces a general context distillation method, not specifically an autonomous agent architecture.
V-SHARE: Provable Convergence Guarantees for Advantage-based Algorithms Provides theoretical convergence guarantees and finite-sample bounds for advantage-based value sharing in reinforcement learning algorithms. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Presents theoretical RL convergence proofs, not LLM-based agent planning or tool use.
V-SHARE: Provable Convergence Guarantees for Advantage-based Algorithms Provides theoretical convergence guarantees and finite-sample bounds for advantage-based value sharing in reinforcement learning algorithms. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Theoretical RL value learning, not model compression, distillation, or efficient serving.
V-SHARE: Provable Convergence Guarantees for Advantage-based Algorithms Provides theoretical convergence guarantees and finite-sample bounds for advantage-based value sharing in reinforcement learning algorithms. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Pure RL algorithmic theory without application to LLM safety, guardrails, or alignment risks.
V-SHARE: Provable Convergence Guarantees for Advantage-based Algorithms Provides theoretical convergence guarantees and finite-sample bounds for advantage-based value sharing in reinforcement learning algorithms. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on theoretical RL convergence analysis, not LLM benchmark evaluation or judge reliability.
Incentivizing Tool-Use Reinforcement Learning for Post-Arrival Emergency Care Prediction Integrates reinforcement learning and tabular model guidance to help language models select features and predict emergency department clinical risk. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Evaluates a predictive clinical agent, not developing novel LLM benchmarks or evaluation methods.
Incentivizing Tool-Use Reinforcement Learning for Post-Arrival Emergency Care Prediction Integrates reinforcement learning and tabular model guidance to help language models select features and predict emergency department clinical risk. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on clinical risk prediction accuracy rather than AI safety, red-teaming, or guardrails.
X-Fly: Learning to Reach and Stop in Full-Chunk Aerial Vision-Language Navigation Proposes a training framework enabling aerial vision-language agents to navigate in full action chunks and stop reliably at destinations. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Addresses aerial drone navigation, not industrial manufacturing inspection or defect detection.
X-Fly: Learning to Reach and Stop in Full-Chunk Aerial Vision-Language Navigation Proposes a training framework enabling aerial vision-language agents to navigate in full action chunks and stop reliably at destinations. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on robotic navigation control rather than system safety, jailbreaks, or guardrails.
X-Fly: Learning to Reach and Stop in Full-Chunk Aerial Vision-Language Navigation Proposes a training framework enabling aerial vision-language agents to navigate in full action chunks and stop reliably at destinations. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Proposes an aerial navigation policy rather than general benchmark evaluation methodologies.
Full-Duplex Interaction with Speech, Expressions and Gestures Presents a full-duplex conversational model that generates real-time speech alongside synchronized facial expressions and body gestures. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Studies conversational gesture synthesis rather than AI safety, hallucination mitigation, or guardrails.
Full-Duplex Interaction with Speech, Expressions and Gestures Presents a full-duplex conversational model that generates real-time speech alongside synchronized facial expressions and body gestures. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on conversational dialogue and co-speech synthesis, not autonomous agent tool use.
Full-Duplex Interaction with Speech, Expressions and Gestures Presents a full-duplex conversational model that generates real-time speech alongside synchronized facial expressions and body gestures. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Introduces a model pipeline and specialized metrics, not general LLM evaluation or benchmark critique.
Full-Duplex Interaction with Speech, Expressions and Gestures Presents a full-duplex conversational model that generates real-time speech alongside synchronized facial expressions and body gestures. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Covers face and gesture animation, not industrial defect detection or visual quality control.
Can Visual Generation Expand What Unified Multimodal Models Know? Investigates whether training multimodal models on visual generation tasks transfers accessible factual knowledge into text question answering. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Explores multimodal knowledge grounding rather than deployment safety, red-teaming, or jailbreaking.
Can Visual Generation Expand What Unified Multimodal Models Know? Investigates whether training multimodal models on visual generation tasks transfers accessible factual knowledge into text question answering. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Investigates cross-modal knowledge transfer, not multi-step tool planning or agent execution.
Can Visual Generation Expand What Unified Multimodal Models Know? Investigates whether training multimodal models on visual generation tasks transfers accessible factual knowledge into text question answering. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Uses probing to analyze representation rather than benchmarking general LLM capability or trustworthiness.
Can Visual Generation Expand What Unified Multimodal Models Know? Investigates whether training multimodal models on visual generation tasks transfers accessible factual knowledge into text question answering. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Studies unified visual generation, not industrial manufacturing inspection or defect detection.
Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Introduces a framework where a small model drafts tool-use trajectories and a larger model verifies them in single prefill passes. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Introduces an agent framework and evaluates on existing benchmarks, not novel benchmark design.
Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Introduces a framework where a small model drafts tool-use trajectories and a larger model verifies them in single prefill passes. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Focuses on language and tool interaction without image, video, or multimodal inputs.
Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Introduces a framework where a small model drafts tool-use trajectories and a larger model verifies them in single prefill passes. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Focuses on agent tool-execution efficiency rather than retrieval-augmented generation pipelines.
Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Introduces a framework where a small model drafts tool-use trajectories and a larger model verifies them in single prefill passes. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Addresses latency and speculative execution rather than deployment safety, alignment, or guardrails.
Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Introduces a framework where a small model drafts tool-use trajectories and a larger model verifies them in single prefill passes. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Evaluates general tool-use agents, not specialized code generation or software engineering repair.
Ontology Refinement with Embeddings and LLMs: A Case Study on CASTON Combines domain embeddings with language models to detect and repair incorrect subsumption axioms in industrial knowledge ontologies. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Domain-specific ontology repair rather than model safety, jailbreak robustness, or guardrails.
Ontology Refinement with Embeddings and LLMs: A Case Study on CASTON Combines domain embeddings with language models to detect and repair incorrect subsumption axioms in industrial knowledge ontologies. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Case study on ontology repair, not a benchmark or evaluation method for LLMs.
Ontology Refinement with Embeddings and LLMs: A Case Study on CASTON Combines domain embeddings with language models to detect and repair incorrect subsumption axioms in industrial knowledge ontologies. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on ontology refinement using embeddings and prompt suggestions, not autonomous agents.
Can LLMs Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation Introduces a benchmark and evaluation to test whether language models can successfully generate coherent nonlinear narratives. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not involve external document retrieval, grounding, or retrieval-augmented generation systems.
Can LLMs Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation Introduces a benchmark and evaluation to test whether language models can successfully generate coherent nonlinear narratives. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on narrative generation, not tool use or autonomous multi-step agent execution.
Can LLMs Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation Introduces a benchmark and evaluation to test whether language models can successfully generate coherent nonlinear narratives. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Covers text-only storytelling without any multimodal, image, video, or speech components.
Can LLMs Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation Introduces a benchmark and evaluation to test whether language models can successfully generate coherent nonlinear narratives. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Deals with natural language storytelling, not code generation, repair, or software engineering.
Can LLMs Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation Introduces a benchmark and evaluation to test whether language models can successfully generate coherent nonlinear narratives. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Evaluates narrative ordering rather than AI safety, jailbreaks, guardrails, or deployment risks.
Can LLMs Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation Introduces a benchmark and evaluation to test whether language models can successfully generate coherent nonlinear narratives. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Evaluates existing models on story generation without proposing distillation, pruning, or efficiency techniques.
Tool-Augmented Large Language Models for Reliable Medical Calculation from Patient Notes Uses tool-augmented language models to extract clinical variables from patient notes and reliably execute deterministic medical calculation formulas. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Uses an existing benchmark (MedCalc-Bench) rather than contributing novel LLM evaluation methodology.
AI-Driven Disaster Response and Displacement Monitoring Surveys the application of artificial intelligence methods to humanitarian disaster response and displacement monitoring. Peer-reviewed	AI agents via arxiv · 17 Oct, 06:17	Not relevant High-level overview paper, not focused on autonomous LLM agent execution or tool use.
AI-Driven Disaster Response and Displacement Monitoring Surveys the application of artificial intelligence methods to humanitarian disaster response and displacement monitoring. Peer-reviewed	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Discusses disaster and satellite monitoring, not industrial defect detection or manufacturing inspection.
AI-Driven Disaster Response and Displacement Monitoring Surveys the application of artificial intelligence methods to humanitarian disaster response and displacement monitoring. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Survey overview, not introducing or analyzing LLM benchmarks or evaluation methods.
AI-Driven Disaster Response and Displacement Monitoring Surveys the application of artificial intelligence methods to humanitarian disaster response and displacement monitoring. Peer-reviewed	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Broad humanitarian survey, not investigating multimodal models or vision-language architectures.

5,281–5,340 of 17,313← Previous Next →