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
Evaluating Model Capabilities After Benchmark Saturation Tests whether transcripts of correctly solved tasks from saturated benchmarks can predict model performance on unseen harder evaluations. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Focuses on post-saturation benchmark evaluation, not model efficiency or compression.
Evaluating Model Capabilities After Benchmark Saturation Tests whether transcripts of correctly solved tasks from saturated benchmarks can predict model performance on unseen harder evaluations. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Studies evaluation methodologies on saturated benchmarks rather than autonomous agent systems.
Value-Informed Schedule Trimming for Accelerated Sequential Monte Carlo Guidance in Discrete Diffusion The paper introduces a training-free schedule optimization method for accelerating sequential Monte Carlo guidance in discrete diffusion models. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Presents a diffusion sampling schedule rather than autonomous planning agents.
Value-Informed Schedule Trimming for Accelerated Sequential Monte Carlo Guidance in Discrete Diffusion The paper introduces a training-free schedule optimization method for accelerating sequential Monte Carlo guidance in discrete diffusion models. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Addresses diffusion sampling efficiency rather than AI deployment safety or guardrails.
Value-Informed Schedule Trimming for Accelerated Sequential Monte Carlo Guidance in Discrete Diffusion The paper introduces a training-free schedule optimization method for accelerating sequential Monte Carlo guidance in discrete diffusion models. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Develops a guidance algorithm rather than an LLM evaluation framework or benchmark.
AlphaDiana: A Measurement System for Agents in Verifiable Reasoning Tasks The paper introduces a modular measurement system to evaluate how agent frameworks, tools, and memory impact verifiable reasoning tasks. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on agent accuracy and harness benchmarking rather than safety, guardrails, or red-teaming.
AlphaDiana: A Measurement System for Agents in Verifiable Reasoning Tasks The paper introduces a modular measurement system to evaluate how agent frameworks, tools, and memory impact verifiable reasoning tasks. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Evaluates token usage across agent harnesses, not model compression or efficient serving.
When Correctness Feedback Erases Evidence: Phase-Separated Completeness Auditing in Verus The paper demonstrates how correctness feedback undermines formal specification auditing and proposes a phase-separated alternative for Verus. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Examines specification auditing dynamics rather than general LLM evaluation benchmarks.
When Correctness Feedback Erases Evidence: Phase-Separated Completeness Auditing in Verus The paper demonstrates how correctness feedback undermines formal specification auditing and proposes a phase-separated alternative for Verus. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Studies formal contract verification and auditing rather than autonomous agent systems.
Counterfactual Neighborhood Risk Fields: Evidence-Grounded Selective Safety for Vision-Language Models The paper proposes a selective safety framework for vision-language models using counterfactual pairs and representation density estimation. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Proposes a safety defense mechanism rather than establishing a general evaluation benchmark.
Counterfactual Neighborhood Risk Fields: Evidence-Grounded Selective Safety for Vision-Language Models The paper proposes a selective safety framework for vision-language models using counterfactual pairs and representation density estimation. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on representation-level safety filtering for VLMs, not multi-step agent workflows.
IneqAgent: Agentic Inequality Proving in Lean with Symbolic Computation The paper introduces an agentic framework combining language model reasoning with symbolic engines for formal inequality proving in Lean4. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on automated theorem proving rather than AI safety, alignment, or guardrails.
IneqAgent: Agentic Inequality Proving in Lean with Symbolic Computation The paper introduces an agentic framework combining language model reasoning with symbolic engines for formal inequality proving in Lean4. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Applies existing math benchmarks to evaluate a theorem-proving system rather than benchmarking methods.
L2G-SR: Learning to Guide LLM-based Symbolic Regression The paper presents a framework that uses structural tree edit distance to guide language models in symbolic regression. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Uses an existing benchmark to test a regression method, rather than proposing evaluation standards.
L2G-SR: Learning to Guide LLM-based Symbolic Regression The paper presents a framework that uses structural tree edit distance to guide language models in symbolic regression. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Addresses mathematical formula discovery rather than deployment safety, alignment, or guardrails.
L2G-SR: Learning to Guide LLM-based Symbolic Regression The paper presents a framework that uses structural tree edit distance to guide language models in symbolic regression. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on symbolic regression algorithms rather than multi-step autonomous agents.
Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans The paper introduces an expert-guided benchmark to measure the impact of generative language models on user wellbeing. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Evaluates multi-turn chat interactions rather than autonomous tool-using agents.
WorldCast: Exploring Intelligent Visual Generation in World State Prediction The paper introduces a hierarchical benchmark for evaluating visual generative models on multi-step world state prediction tasks. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Focuses on general open-world visual generation, not industrial manufacturing defect detection.
WorldCast: Exploring Intelligent Visual Generation in World State Prediction The paper introduces a hierarchical benchmark for evaluating visual generative models on multi-step world state prediction tasks. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on visual generation performance rather than AI safety or guardrails.
WorldCast: Exploring Intelligent Visual Generation in World State Prediction The paper introduces a hierarchical benchmark for evaluating visual generative models on multi-step world state prediction tasks. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Evaluates visual generation models rather than multi-step tool-using language agents.
Interpreting Language Model Parameters The paper introduces a parameter decomposition method to interpret the mechanistic subcomponents of language model weights. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Investigates mechanistic interpretability of model weights, not autonomous agent systems.
Interpreting Language Model Parameters The paper introduces a parameter decomposition method to interpret the mechanistic subcomponents of language model weights. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Focuses on interpretability and parameter decomposition rather than model compression for deployment.
Interpreting Language Model Parameters The paper introduces a parameter decomposition method to interpret the mechanistic subcomponents of language model weights. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Studies parameter decomposition rather than benchmark saturation, LLM-as-a-judge, or evaluation protocols.
Optimal Hydra: Model Heterogeneous Federated Learning for LLMs using Data-Informed Hydra-Nets The paper proposes a federated learning framework that tailors sub-network sizes and architectures to heterogeneous client hardware and data. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on federated optimization rather than AI safety, hallucination, or guardrails.
Optimal Hydra: Model Heterogeneous Federated Learning for LLMs using Data-Informed Hydra-Nets The paper proposes a federated learning framework that tailors sub-network sizes and architectures to heterogeneous client hardware and data. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Proposes a federated training algorithm rather than benchmarking or evaluation protocols.
Optimal Hydra: Model Heterogeneous Federated Learning for LLMs using Data-Informed Hydra-Nets The paper proposes a federated learning framework that tailors sub-network sizes and architectures to heterogeneous client hardware and data. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on federated learning architecture rather than multi-step autonomous agent systems.
Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models The paper identifies an internal activation direction in spoken language models to control and transfer conversational turn-taking styles. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not address retrieval-augmented generation or grounding from external databases.
Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models The paper identifies an internal activation direction in spoken language models to control and transfer conversational turn-taking styles. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Controls conversational turn-taking style rather than safety, guardrails, or adversarial threats.
Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models The paper identifies an internal activation direction in spoken language models to control and transfer conversational turn-taking styles. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Studies activation steering vectors, not model compression or efficient serving.
Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models The paper identifies an internal activation direction in spoken language models to control and transfer conversational turn-taking styles. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on steering spoken conversational style rather than LLM benchmarking methodology.
Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models The paper identifies an internal activation direction in spoken language models to control and transfer conversational turn-taking styles. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Investigates full-duplex spoken models rather than autonomous multi-step planning agents.
Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models The paper identifies an internal activation direction in spoken language models to control and transfer conversational turn-taking styles. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Does not involve code generation, software engineering, or program repair.
When Reality Outpaces Cognition: A New Scaling Principle for Reliable Intelligence The paper presents a metric and benchmark measuring how quickly environment dynamics can change before agent decision-making fails. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Focuses on biological-inspired cognition speed rather than model quantization or lightweight deployment.
When Reality Outpaces Cognition: A New Scaling Principle for Reliable Intelligence The paper presents a metric and benchmark measuring how quickly environment dynamics can change before agent decision-making fails. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Investigates theoretical reinforcement learning dynamics rather than language model agent systems.
When Reality Outpaces Cognition: A New Scaling Principle for Reliable Intelligence The paper presents a metric and benchmark measuring how quickly environment dynamics can change before agent decision-making fails. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Examines general dynamic environments, not industrial manufacturing or visual defect inspection.
When Reality Outpaces Cognition: A New Scaling Principle for Reliable Intelligence The paper presents a metric and benchmark measuring how quickly environment dynamics can change before agent decision-making fails. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Proposes a conceptual scaling metric rather than standard LLM evaluation frameworks.
Verdict: Encode Once, Decide Many The paper introduces an encoder architecture that processes input once to answer multiple structured classification queries simultaneously. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Applies existing benchmarks to evaluate a model architecture rather than studying benchmarks themselves.
Verdict: Encode Once, Decide Many The paper introduces an encoder architecture that processes input once to answer multiple structured classification queries simultaneously. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Introduces an efficient single-pass classifier architecture, not an autonomous agent.
Verdict: Encode Once, Decide Many The paper introduces an encoder architecture that processes input once to answer multiple structured classification queries simultaneously. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Does not evaluate code generation, software maintenance, or program repair.
Verdict: Encode Once, Decide Many The paper introduces an encoder architecture that processes input once to answer multiple structured classification queries simultaneously. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Operates entirely on text classification without multimodal data.
Verdict: Encode Once, Decide Many The paper introduces an encoder architecture that processes input once to answer multiple structured classification queries simultaneously. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Evaluates multi-question classification rather than document retrieval or RAG systems.
Verdict: Encode Once, Decide Many The paper introduces an encoder architecture that processes input once to answer multiple structured classification queries simultaneously. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on classification efficiency rather than AI safety, guardrails, or robustness.
ARIA: Runtime Inspection and Amendment for Multi-Agent Systems The paper presents a framework and dataset for learning to detect and repair errors during multi-agent system execution. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Focuses on language-based multi-agent execution rather than multimodal models.
ARIA: Runtime Inspection and Amendment for Multi-Agent Systems The paper presents a framework and dataset for learning to detect and repair errors during multi-agent system execution. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Deals with multi-agent error repair rather than retrieval-augmented generation systems.
ARIA: Runtime Inspection and Amendment for Multi-Agent Systems The paper presents a framework and dataset for learning to detect and repair errors during multi-agent system execution. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Does not focus on model compression, quantization, or efficient inference.
ARIA: Runtime Inspection and Amendment for Multi-Agent Systems The paper presents a framework and dataset for learning to detect and repair errors during multi-agent system execution. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Targets general multi-agent execution trajectories rather than code generation or software engineering.
ARIA: Runtime Inspection and Amendment for Multi-Agent Systems The paper presents a framework and dataset for learning to detect and repair errors during multi-agent system execution. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Introduces a dataset of repair trajectories for training, not a general evaluation benchmark.
ReasonUni: Reasoning-Augmented Reconstruction Advances Unified Multimodal Models The paper introduces a reasoning-augmented reconstructive post-training framework that unifies multimodal understanding and image generation tasks. Preprint	Computer vision for industrial applications via arxiv · 17 Oct, 06:17	Not relevant Studies general image editing and generation, not industrial inspection or defect detection.
ReasonUni: Reasoning-Augmented Reconstruction Advances Unified Multimodal Models The paper introduces a reasoning-augmented reconstructive post-training framework that unifies multimodal understanding and image generation tasks. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Does not address safety, guardrails, hallucination, or adversarial robustness.
ReasonUni: Reasoning-Augmented Reconstruction Advances Unified Multimodal Models The paper introduces a reasoning-augmented reconstructive post-training framework that unifies multimodal understanding and image generation tasks. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Presents a model training framework rather than benchmarking or evaluation methodology.
ReasonUni: Reasoning-Augmented Reconstruction Advances Unified Multimodal Models The paper introduces a reasoning-augmented reconstructive post-training framework that unifies multimodal understanding and image generation tasks. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on unified multimodal generation and editing, not autonomous multi-step agent systems.
Open-world AI: a Tale of Three Requirements The paper formalizes three core architectural requirements for artificial intelligence systems operating in open-world settings. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Theoretical analysis of open-world generalization, not autonomous agent systems.
Open-world AI: a Tale of Three Requirements The paper formalizes three core architectural requirements for artificial intelligence systems operating in open-world settings. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Studies generalization theory rather than benchmarking or evaluation methodology.
Open-world AI: a Tale of Three Requirements The paper formalizes three core architectural requirements for artificial intelligence systems operating in open-world settings. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Addresses representation learning theory rather than deployment safety, guardrails, or robustness.
EvolvingBrain: An Adaptive Skill System for Multi-modal Agents Self-Evolution in Open Worlds The paper proposes an adaptive skill framework that extracts patterns from trajectories to enable multi-modal agents to self-evolve. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on agent architecture rather than evaluating general benchmark reliability.
EvolvingBrain: An Adaptive Skill System for Multi-modal Agents Self-Evolution in Open Worlds The paper proposes an adaptive skill framework that extracts patterns from trajectories to enable multi-modal agents to self-evolve. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Targets agent capability and skill acquisition rather than safety, guardrails, or alignment.
When Reasoning Traces Conflict with Evidence: How Answer Configuration Shapes Evidence Following The paper analyzes how answering via a separate model call improves evidence adherence when reasoning traces conflict with provided context. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Does not explore model compression, distillation, quantization, or efficient serving.
When Reasoning Traces Conflict with Evidence: How Answer Configuration Shapes Evidence Following The paper analyzes how answering via a separate model call improves evidence adherence when reasoning traces conflict with provided context. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on trace answering setups rather than multi-step agent planning or tools.
When Reasoning Traces Conflict with Evidence: How Answer Configuration Shapes Evidence Following The paper analyzes how answering via a separate model call improves evidence adherence when reasoning traces conflict with provided context. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Operates strictly on text QA benchmarks without multimodal inputs.
When Reasoning Traces Conflict with Evidence: How Answer Configuration Shapes Evidence Following The paper analyzes how answering via a separate model call improves evidence adherence when reasoning traces conflict with provided context. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Analyzes evidence-following mechanisms rather than establishing a general evaluation benchmark.

4,861–4,920 of 17,313← Previous Next →