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
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. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on document entity extraction without retrieval-augmented generation systems.
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. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Evaluates entity extraction tasks, not multi-step autonomous planning agents.
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. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Deals strictly with text-based entity extraction from documents.
OrionMoE-v1.0: A Mixture-of-Experts Foundation Model for Tabular In-Context Learning The paper introduces a compact tabular foundation model combining distribution-aware representations with dataset-conditioned sparse mixture-of-experts computation. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on tabular foundation models, not multi-step tool-using agents.
OrionMoE-v1.0: A Mixture-of-Experts Foundation Model for Tabular In-Context Learning The paper introduces a compact tabular foundation model combining distribution-aware representations with dataset-conditioned sparse mixture-of-experts computation. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes a model architecture, not a general evaluation benchmark.
OrionMoE-v1.0: A Mixture-of-Experts Foundation Model for Tabular In-Context Learning The paper introduces a compact tabular foundation model combining distribution-aware representations with dataset-conditioned sparse mixture-of-experts computation. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on predictive tabular modeling rather than deployment safety.
Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks The paper analyzes the performance of retrieval-augmented generation architectures on open-domain question answering tasks involving culturally grounded South Asian narratives. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not evaluate jailbreaks, hallucinations, or safety guardrails.
Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks The paper analyzes the performance of retrieval-augmented generation architectures on open-domain question answering tasks involving culturally grounded South Asian narratives. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Empirical analysis of RAG factors, not a benchmark framework.
Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks The paper analyzes the performance of retrieval-augmented generation architectures on open-domain question answering tasks involving culturally grounded South Asian narratives. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on text question answering, not multimodal models.
Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks The paper analyzes the performance of retrieval-augmented generation architectures on open-domain question answering tasks involving culturally grounded South Asian narratives. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model compression or efficient serving.
Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks The paper analyzes the performance of retrieval-augmented generation architectures on open-domain question answering tasks involving culturally grounded South Asian narratives. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on QA retrieval architectures, not multi-step agents.
Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks The paper analyzes the performance of retrieval-augmented generation architectures on open-domain question answering tasks involving culturally grounded South Asian narratives. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Studies open-domain question answering, not software engineering.
SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm The paper presents a recursive self-improvement framework using scientific agent swarms that mutate hypotheses and recombine code based on experimental feedback. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Targets automated scientific research rather than software engineering tools.
SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm The paper presents a recursive self-improvement framework using scientific agent swarms that mutate hypotheses and recombine code based on experimental feedback. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Does not address multimodal language-vision models.
SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm The paper presents a recursive self-improvement framework using scientific agent swarms that mutate hypotheses and recombine code based on experimental feedback. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes an agent system rather than an evaluation benchmark.
SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm The paper presents a recursive self-improvement framework using scientific agent swarms that mutate hypotheses and recombine code based on experimental feedback. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on self-improvement rather than safety or reliability.
SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm The paper presents a recursive self-improvement framework using scientific agent swarms that mutate hypotheses and recombine code based on experimental feedback. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not cover model quantization, pruning, or distillation.
SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm The paper presents a recursive self-improvement framework using scientific agent swarms that mutate hypotheses and recombine code based on experimental feedback. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on agent swarm evolution, not document retrieval.
From Training to Escalation: Forensic Analysis and Local Recreation of RL/RLHF-Induced Privilege Escalation in AI Agents (CVE-2026-65616) Recreates and analyzes infrastructure misconfigurations that enabled privilege escalation and repository bypass during autonomous AI agent evaluations. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Forensic security analysis rather than benchmark methodology.
Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Presents an OCR evaluation metric, not an agent system.
Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on OCR transcription errors, not code generation.
Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Does not address retrieval-augmented generation pipelines.
Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on transcription scoring rather than deployment safety.
Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Evaluates general OCR transcription, not industrial visual inspection.
Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not cover model compression or inference efficiency.
Selective Growing Q-Networks (SGQN): An Analysis of Adaptive Action Grids in Decoupled Q-Learning Implements adaptive action grids in value-based reinforcement learning that selectively refine promising bins while pruning suboptimal actions. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Studies continuous control reinforcement learning, not LLM agents.
Selective Growing Q-Networks (SGQN): An Analysis of Adaptive Action Grids in Decoupled Q-Learning Implements adaptive action grids in value-based reinforcement learning that selectively refine promising bins while pruning suboptimal actions. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on RL action discretization, not deployment safety.
Selective Growing Q-Networks (SGQN): An Analysis of Adaptive Action Grids in Decoupled Q-Learning Implements adaptive action grids in value-based reinforcement learning that selectively refine promising bins while pruning suboptimal actions. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes continuous control RL grids, not LLM evaluation.
Selective Growing Q-Networks (SGQN): An Analysis of Adaptive Action Grids in Decoupled Q-Learning Implements adaptive action grids in value-based reinforcement learning that selectively refine promising bins while pruning suboptimal actions. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on Q-learning grids, not small language models.
SafeAdapt-Poker: Counterfactual Risk Calibration for Budgeted Adaptation under Opponent Shift Introduces a counterfactual risk management method to calibrate adaptive decision-making against shifting opponent behaviors within fixed deviation budgets. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not deal with model compression or efficient serving.
SafeAdapt-Poker: Counterfactual Risk Calibration for Budgeted Adaptation under Opponent Shift Introduces a counterfactual risk management method to calibrate adaptive decision-making against shifting opponent behaviors within fixed deviation budgets. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Poker game theory adaptation, not deployment safety or guardrails.
SafeAdapt-Poker: Counterfactual Risk Calibration for Budgeted Adaptation under Opponent Shift Introduces a counterfactual risk management method to calibrate adaptive decision-making against shifting opponent behaviors within fixed deviation budgets. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents an adaptation algorithm, not an evaluation benchmark.
SafeAdapt-Poker: Counterfactual Risk Calibration for Budgeted Adaptation under Opponent Shift Introduces a counterfactual risk management method to calibrate adaptive decision-making against shifting opponent behaviors within fixed deviation budgets. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on game theory and reinforcement learning in poker.
Quantifying privacy risks in foundation model representations of brain MRI Evaluates patient linkage and attribute inference risks present in brain magnetic resonance imaging representations generated by pretrained foundation models. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Evaluates brain MRI representations, not industrial defect inspection.
Quantifying privacy risks in foundation model representations of brain MRI Evaluates patient linkage and attribute inference risks present in brain magnetic resonance imaging representations generated by pretrained foundation models. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on brain MRI encoders, not multimodal VLMs.
Quantifying privacy risks in foundation model representations of brain MRI Evaluates patient linkage and attribute inference risks present in brain magnetic resonance imaging representations generated by pretrained foundation models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Privacy risk analysis rather than a general LLM benchmark.
Quantifying privacy risks in foundation model representations of brain MRI Evaluates patient linkage and attribute inference risks present in brain magnetic resonance imaging representations generated by pretrained foundation models. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Studies privacy leakage in medical representations, not autonomous agents.
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. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on RL token-level credit assignment, not agent architectures.
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. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses RL optimization rather than safety or alignment guardrails.
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. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents an RL training algorithm rather than an evaluation benchmark.
VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution Uses explicit three-dimensional vascular guidance in an iterative diffusion pipeline to improve anatomical accuracy in computed tomography super-resolution. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not cover AI safety, jailbreaks, or guardrails.
VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution Uses explicit three-dimensional vascular guidance in an iterative diffusion pipeline to improve anatomical accuracy in computed tomography super-resolution. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes a generative method, not an evaluation framework.
VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution Uses explicit three-dimensional vascular guidance in an iterative diffusion pipeline to improve anatomical accuracy in computed tomography super-resolution. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Medical image reconstruction, not multimodal vision-language models.
VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution Uses explicit three-dimensional vascular guidance in an iterative diffusion pipeline to improve anatomical accuracy in computed tomography super-resolution. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Focuses on medical CT scans, not industrial manufacturing inspection.
VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution Uses explicit three-dimensional vascular guidance in an iterative diffusion pipeline to improve anatomical accuracy in computed tomography super-resolution. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address efficient language model serving or compression.
VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution Uses explicit three-dimensional vascular guidance in an iterative diffusion pipeline to improve anatomical accuracy in computed tomography super-resolution. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Medical CT imaging diffusion method, not an agent system.
VASPO: Self-Distillation with Privileged Value Estimation Combines privileged teacher signals with value estimation in policy optimization to provide denser token-level supervision for reasoning tasks. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Reinforcement learning and distillation technique, not agent tool use.
VASPO: Self-Distillation with Privileged Value Estimation Combines privileged teacher signals with value estimation in policy optimization to provide denser token-level supervision for reasoning tasks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses training dynamics rather than deployment safety risks.
VASPO: Self-Distillation with Privileged Value Estimation Combines privileged teacher signals with value estimation in policy optimization to provide denser token-level supervision for reasoning tasks. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses exclusively on text-based mathematical reasoning.
VASPO: Self-Distillation with Privileged Value Estimation Combines privileged teacher signals with value estimation in policy optimization to provide denser token-level supervision for reasoning tasks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes an RL training algorithm, not an evaluation method.
VASPO: Self-Distillation with Privileged Value Estimation Combines privileged teacher signals with value estimation in policy optimization to provide denser token-level supervision for reasoning tasks. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Evaluated on mathematical reasoning benchmarks, not code generation.
VASPO: Self-Distillation with Privileged Value Estimation Combines privileged teacher signals with value estimation in policy optimization to provide denser token-level supervision for reasoning tasks. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Does not involve external retrieval or RAG systems.
The Wealth of Agents: Scaling Agentic AI Across Developers Studies multi-developer agentic systems where independently built specialist agents collaborate to outperform larger systems created by single developers. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Studies agent collaboration frameworks rather than deployment safety.
The Wealth of Agents: Scaling Agentic AI Across Developers Studies multi-developer agentic systems where independently built specialist agents collaborate to outperform larger systems created by single developers. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Evaluates existing benchmarks rather than proposing evaluation methodology.
ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps Presents a graph-based framework with decision certificates to determine when multimodal agents should acquire visual evidence under cost constraints. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on cost-sensitive acquisition rather than AI safety.
ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps Presents a graph-based framework with decision certificates to determine when multimodal agents should acquire visual evidence under cost constraints. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not study model compression or efficient serving architectures.
ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps Presents a graph-based framework with decision certificates to determine when multimodal agents should acquire visual evidence under cost constraints. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents an acquisition algorithm rather than an evaluation benchmark.
ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps Presents a graph-based framework with decision certificates to determine when multimodal agents should acquire visual evidence under cost constraints. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Theoretical active evidence framework, not industrial visual inspection.
PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis Introduces a decomposed rubric synthesis method to create training data for open multimodal models that generate visual code artifacts. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on code intelligence and artifact generation, not agents.
PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis Introduces a decomposed rubric synthesis method to create training data for open multimodal models that generate visual code artifacts. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Does not cover document retrieval or RAG architectures.

8,041–8,100 of 17,313← Previous Next →