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
Agreement Without Alignment: Label-Free Rubric Clarification Increases LLM-Judge Consensus but Not Agreement with Humans Evaluates a label-free rubric clarification method, finding it increases consensus among LLM judges without improving agreement with human evaluations. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Does not address retrieval-augmented generation, chunking, or knowledge-base search.
Agreement Without Alignment: Label-Free Rubric Clarification Increases LLM-Judge Consensus but Not Agreement with Humans Evaluates a label-free rubric clarification method, finding it increases consensus among LLM judges without improving agreement with human evaluations. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not address model efficiency, distillation, quantization, or compact architectures.
Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning Introduces ChemDiffAgent, applying masked diffusion to plan tool use and reasoning trajectories bidirectionally for chemical problem-solving. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not investigate small language models, quantization, pruning, or efficient inference techniques.
Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning Introduces ChemDiffAgent, applying masked diffusion to plan tool use and reasoning trajectories bidirectionally for chemical problem-solving. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not investigate safety risks, prompt injections, guardrails, or alignment vulnerabilities.
Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning Introduces ChemDiffAgent, applying masked diffusion to plan tool use and reasoning trajectories bidirectionally for chemical problem-solving. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes a new agent architecture rather than evaluating benchmarks or LLM-as-a-judge schemes.
AlphaNIST: Emergent Assembly Skills from Binary Task Rewards Evaluates whether robotic manipulation agents can learn complex multi-step assembly skills purely from sparse binary task rewards. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not study LLM safety, guardrails, hallucinations, or deployment red-teaming.
AlphaNIST: Emergent Assembly Skills from Binary Task Rewards Evaluates whether robotic manipulation agents can learn complex multi-step assembly skills purely from sparse binary task rewards. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on reinforcement learning for robotic manipulation, not LLM-based autonomous planning agents.
AlphaNIST: Emergent Assembly Skills from Binary Task Rewards Evaluates whether robotic manipulation agents can learn complex multi-step assembly skills purely from sparse binary task rewards. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Examines RL policy learning rather than testing or benchmarking LLM evaluation methods.
AgentSWE: Can Coding Agents Build the Agent You Actually Want? Introduces AgentSWE, a 25-task benchmark evaluating whether coding agents can build, edit, and optimize other software agents against user requirements. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not study model compression, distillation, quantization, or efficient on-device serving.
ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents Introduces ProgramGym, a framework providing verifiable sandbox environments and black-box test feedback for training long-horizon program reproduction agents. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Applies solely to source code generation and software environments without multimodal or vision data.
ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents Introduces ProgramGym, a framework providing verifiable sandbox environments and black-box test feedback for training long-horizon program reproduction agents. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on software engineering capability rather than AI safety guardrails, alignment, or jailbreaks.
ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents Introduces ProgramGym, a framework providing verifiable sandbox environments and black-box test feedback for training long-horizon program reproduction agents. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Focuses on code reproduction environments rather than model compression or inference efficiency.
ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents Introduces ProgramGym, a framework providing verifiable sandbox environments and black-box test feedback for training long-horizon program reproduction agents. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Focuses on program exploration and verifiable environments rather than retrieval-augmented generation architectures.
RegulaTutor: A Theory-Informed Framework for Long-Term Personalized Learning Presents RegulaTutor, a personalized tutoring framework that grounds instructional policies in explicit cognitive and self-regulated learning theories. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses pedagogical instructional policies, not AI safety risks, guardrails, or jailbreaks.
RegulaTutor: A Theory-Informed Framework for Long-Term Personalized Learning Presents RegulaTutor, a personalized tutoring framework that grounds instructional policies in explicit cognitive and self-regulated learning theories. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents a personalized tutoring framework rather than a general LLM evaluation benchmark.
In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validation Develops a probabilistic transformer meta-learning architecture to perform zero-shot system identification directly from long training sequence contexts. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Applies meta-learning to dynamical systems rather than evaluating LLMs on standardized tasks.
In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validation Develops a probabilistic transformer meta-learning architecture to perform zero-shot system identification directly from long training sequence contexts. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Studies in-context learning for classical dynamical system identification, not language model agents.
In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validation Develops a probabilistic transformer meta-learning architecture to perform zero-shot system identification directly from long training sequence contexts. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Addresses meta-learning for dynamical systems rather than model compression, distillation, or lightweight serving.
In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validation Develops a probabilistic transformer meta-learning architecture to perform zero-shot system identification directly from long training sequence contexts. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on control theory system identification rather than LLM safety, guardrails, or reliability risks.
Web$hell: A Unified Interface for Agentic Web Access Proposes Web$hell, a unified interface combining broad information gathering, browser interactions, and evidence management for web-based agents. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on agent interface efficiency and tool execution rather than deployment safety or guardrails.
Web$hell: A Unified Interface for Agentic Web Access Proposes Web$hell, a unified interface combining broad information gathering, browser interactions, and evidence management for web-based agents. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Introduces an agent web-interaction interface rather than a standardized evaluation benchmark.
CNN-Transformer with Sigmoid Label for High-Resolution Sub-Degree DOA Estimation Combines a CNN-Transformer network with a continuous sigmoid label encoding to overcome grid quantization errors in direction-of-arrival estimation. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Covers signal processing label encodings, not AI safety, robustness, jailbreaks, or guardrails.
CNN-Transformer with Sigmoid Label for High-Resolution Sub-Degree DOA Estimation Combines a CNN-Transformer network with a continuous sigmoid label encoding to overcome grid quantization errors in direction-of-arrival estimation. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Addresses signal processing direction-of-arrival estimation using CNN-Transformers, not LLM agents.
CNN-Transformer with Sigmoid Label for High-Resolution Sub-Degree DOA Estimation Combines a CNN-Transformer network with a continuous sigmoid label encoding to overcome grid quantization errors in direction-of-arrival estimation. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Focuses on signal processing DOA accuracy rather than LLM evaluation or benchmark critique.
Local Diagnostics of Learned Event Representations for Anomaly Detection at the HL-LHC Evaluates neural embedding and diagnostic strategies to interpret detected anomalies in high-luminosity collider particle physics data. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on high-energy particle physics collision representations, not language model agents.
Local Diagnostics of Learned Event Representations for Anomaly Detection at the HL-LHC Evaluates neural embedding and diagnostic strategies to interpret detected anomalies in high-luminosity collider particle physics data. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Studies physics collision anomaly detection rather than evaluating general language models.
Local Diagnostics of Learned Event Representations for Anomaly Detection at the HL-LHC Evaluates neural embedding and diagnostic strategies to interpret detected anomalies in high-luminosity collider particle physics data. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses particle physics anomaly detection rather than AI deployment safety, guardrails, or alignment.
Rollout-Time Intervention: Unlocking Agentic Reinforcement Learning Beyond Policy Limits Proposes Rollout-Time Intervention to convert failed agent trajectories into successful off-policy training signals during reinforcement learning. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on improving agent task performance via RL interventions rather than deployment safety or guardrails.
Rollout-Time Intervention: Unlocking Agentic Reinforcement Learning Beyond Policy Limits Proposes Rollout-Time Intervention to convert failed agent trajectories into successful off-policy training signals during reinforcement learning. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents an RL training algorithm for agents rather than proposing a benchmark or evaluation methodology.
Collective Autonomous Vehicle Routing Needs Social Awareness for Coexistence with Human Drivers Incorporates marginal-cost rewards into multi-agent reinforcement learning to reduce autonomous vehicle travel times without worsening human driver delays. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Studies traffic routing optimization algorithms rather than language model evaluation or benchmarks.
Collective Autonomous Vehicle Routing Needs Social Awareness for Coexistence with Human Drivers Incorporates marginal-cost rewards into multi-agent reinforcement learning to reduce autonomous vehicle travel times without worsening human driver delays. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Deals with traffic congestion externalities rather than language model deployment safety, red-teaming, or guardrails.
Collective Autonomous Vehicle Routing Needs Social Awareness for Coexistence with Human Drivers Incorporates marginal-cost rewards into multi-agent reinforcement learning to reduce autonomous vehicle travel times without worsening human driver delays. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Investigates traditional multi-agent reinforcement learning for vehicle routing, not language model agents.
Agentic Large Language Models Presents a textbook providing an overview of the concepts, architectures, and interactions of agentic large language models. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Presents a general survey of agentic models rather than a targeted safety, security, or robustness study.
Agentic Large Language Models Presents a textbook providing an overview of the concepts, architectures, and interactions of agentic large language models. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Acts as a broad book on agentic LLMs rather than focusing on evaluation methodologies or benchmarks.
SqueezeBench: Saturation-Resilient Evaluation of Mathematical Capability Introduces SqueezeBench, a saturation-resilient benchmark for formal mathematical reasoning where new accepted proofs dynamically raise the target bounds. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Studies mathematical capability benchmarking, not model compression, distillation, or efficient serving.
SqueezeBench: Saturation-Resilient Evaluation of Mathematical Capability Introduces SqueezeBench, a saturation-resilient benchmark for formal mathematical reasoning where new accepted proofs dynamically raise the target bounds. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on Lean-verified mathematical evaluation rather than LLM agent planning or tool actions.
SqueezeBench: Saturation-Resilient Evaluation of Mathematical Capability Introduces SqueezeBench, a saturation-resilient benchmark for formal mathematical reasoning where new accepted proofs dynamically raise the target bounds. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses benchmark saturation in formal mathematics, not AI safety, alignment, or guardrails.
When MLLMs miss what humans read: diagnosing and learning to read squint text Investigates multimodal LLM failures on squint-readable low-frequency text and provides a diagnostic benchmark alongside training interventions. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Studies visual text reading deficiencies in vision-language models, not agent planning or tool use.
When MLLMs miss what humans read: diagnosing and learning to read squint text Investigates multimodal LLM failures on squint-readable low-frequency text and provides a diagnostic benchmark alongside training interventions. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Evaluates reading stylized text in moderation images, not industrial inspection or manufacturing anomaly detection.
A Local Agentic AI Framework to Generate Edge Case Scenarios for Safety Assurance Presents a local agentic framework using smaller language models to automatically synthesize edge-case testing scenarios for autonomous systems. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Focuses on scenario generation for autonomous system testing, not computer vision defect inspection.
Branching On-Policy Distillation Introduces BranchOPD, which branches student rollouts from shared prefixes at uncertain decision points to improve on-policy distillation efficiency. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Introduces a knowledge distillation technique rather than benchmarking methodology or evaluation critique.
Branching On-Policy Distillation Introduces BranchOPD, which branches student rollouts from shared prefixes at uncertain decision points to improve on-policy distillation efficiency. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on an on-policy distillation training method rather than agent planning or multi-agent architectures.
Branching On-Policy Distillation Introduces BranchOPD, which branches student rollouts from shared prefixes at uncertain decision points to improve on-policy distillation efficiency. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Optimizes distillation training efficiency without studying deployment safety, red-teaming, or guardrails.
Humanize MLE Agents: Multi-Agent Session Alternation Beyond Single-Agent Limits Presents a framework alternating multi-agent sessions across context resets to overcome reasoning degradation in long-horizon machine learning engineering. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses context degradation and agent session management, not safety guardrails or deployment risk mitigation.
Humanize MLE Agents: Multi-Agent Session Alternation Beyond Single-Agent Limits Presents a framework alternating multi-agent sessions across context resets to overcome reasoning degradation in long-horizon machine learning engineering. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes an execution framework for ML engineering agents rather than a standardized evaluation benchmark.
Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning Presents Explanatory World Models, using language model causal explanations to redistribute reinforcement learning rewards across distant actions. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Applies language models to RL credit assignment in recommendations, not retrieval-augmented generation.
Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning Presents Explanatory World Models, using language model causal explanations to redistribute reinforcement learning rewards across distant actions. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on reinforcement learning credit assignment for recommender systems rather than LLM agent workflows.
Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning Presents Explanatory World Models, using language model causal explanations to redistribute reinforcement learning rewards across distant actions. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Addresses RL sample efficiency in recommenders rather than small language models or model compression.
Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning Presents Explanatory World Models, using language model causal explanations to redistribute reinforcement learning rewards across distant actions. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on long-horizon RL optimization, not deployment safety, guardrails, or hallucination prevention.
Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning Presents Explanatory World Models, using language model causal explanations to redistribute reinforcement learning rewards across distant actions. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes an algorithmic credit assignment framework rather than benchmark analysis or evaluation methodology.
Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis Proposes PRISM, a multimodal clinical prognostic model combining patient records with retrieved population-level embedding memory banks. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Applies multimodal methods to clinical electronic health records, not to code generation or software engineering.
Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis Proposes PRISM, a multimodal clinical prognostic model combining patient records with retrieved population-level embedding memory banks. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Presents a diagnostic clinical model without addressing safety guardrails, jailbreaks, or hallucination mitigation.
Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis Proposes PRISM, a multimodal clinical prognostic model combining patient records with retrieved population-level embedding memory banks. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Introduces a predictive multimodal clinical model rather than an interactive multi-step planning agent.
Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis Proposes PRISM, a multimodal clinical prognostic model combining patient records with retrieved population-level embedding memory banks. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Focuses on medical multimodal integration rather than model compression, quantization, or serving efficiency.
Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis Proposes PRISM, a multimodal clinical prognostic model combining patient records with retrieved population-level embedding memory banks. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Evaluates a clinical prognosis architecture on existing benchmarks rather than proposing general LLM benchmark methods.
RESCUE: Is This Alarm Worth Acting On? Counterfactual Evaluation of VLA Runtime Monitors Presents RESCUE, a counterfactual simulation framework to evaluate whether runtime failure monitor alarms enable beneficial robot interventions. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Focuses on embodied robotic manipulation policy monitors rather than industrial inspection or manufacturing computer vision.
Causal Effect Identifiability in Causal Abstractions: A Strict Hierarchy of Criteria Formalizes a theoretical hierarchy of identifiability criteria for estimating causal effects across equivalence classes of causal graphs. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Addresses theoretical causal identifiability mathematics rather than evaluating language models.
Causal Effect Identifiability in Causal Abstractions: A Strict Hierarchy of Criteria Formalizes a theoretical hierarchy of identifiability criteria for estimating causal effects across equivalence classes of causal graphs. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on theoretical causality rather than LLM safety, guardrails, hallucination, or deployment risks.
Causal Effect Identifiability in Causal Abstractions: A Strict Hierarchy of Criteria Formalizes a theoretical hierarchy of identifiability criteria for estimating causal effects across equivalence classes of causal graphs. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses purely on theoretical causal graph abstraction criteria without language model agents or planning.
Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation Proposes Road MOPD, using non-routed teachers as cross-domain critics to resolve gradient conflicts in multi-teacher on-policy distillation. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses optimization gradient conflicts in distillation, not safety, guardrails, jailbreaks, or deployment monitoring.

13,561–13,620 of 13,754← Previous Next →