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
Safe-FCL: Continual Risk Awareness and Fine-Grained Safety Aggregation for Federated Continual Learning of LLMs The paper proposes a federated learning framework with risk-aware aggregation to prevent malicious updates from compromising global model safety. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on federated safety rather than model efficiency or pruning.
Empirical Boundaries of Reservoir Computing for Few-Shot Sequence Pattern Tasks The paper empirically analyzes echo state network variants on synthetic sequence-to-sequence pattern tasks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on synthetic sequence limits, not safety or guardrails.
Empirical Boundaries of Reservoir Computing for Few-Shot Sequence Pattern Tasks The paper empirically analyzes echo state network variants on synthetic sequence-to-sequence pattern tasks. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Evaluates Echo State Networks on synthetic sequences, not agents.
Empirical Boundaries of Reservoir Computing for Few-Shot Sequence Pattern Tasks The paper empirically analyzes echo state network variants on synthetic sequence-to-sequence pattern tasks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Studies recurrent neural net limits rather than LLM benchmarks.
A Strong Harness Narrows the Model Gap: A Comparative Evaluation for Legislative Document-to-XML Conversion The paper evaluates harnesses for legislative document conversion, showing structured multi-agent pipelines narrow performance gaps between models. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on agent harness orchestration rather than model compression.
A Strong Harness Narrows the Model Gap: A Comparative Evaluation for Legislative Document-to-XML Conversion The paper evaluates harnesses for legislative document conversion, showing structured multi-agent pipelines narrow performance gaps between models. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Applies harnesses to document-to-XML conversion, not software engineering.
A Strong Harness Narrows the Model Gap: A Comparative Evaluation for Legislative Document-to-XML Conversion The paper evaluates harnesses for legislative document conversion, showing structured multi-agent pipelines narrow performance gaps between models. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Processes legislative text documents without image, audio, or video modalities.
A Strong Harness Narrows the Model Gap: A Comparative Evaluation for Legislative Document-to-XML Conversion The paper evaluates harnesses for legislative document conversion, showing structured multi-agent pipelines narrow performance gaps between models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Evaluates harness structuring for XML parsing, not retrieval systems.
What Coding Agents Leave Unsaid: Measuring Task-State Gaps Between Execution and Reports The paper measures the information gap between coding agents' internal execution logs and their user-facing progress reports. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on agent oversight gaps rather than general benchmarking.
Provable Transformer Context Compression for In-Context Reinforcement Learning The paper analyzes the minimum context size required for transformer models to perform in-context reinforcement learning without parameter updates. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Theoretical context compression in RL, not multi-step agent systems.
Provable Transformer Context Compression for In-Context Reinforcement Learning The paper analyzes the minimum context size required for transformer models to perform in-context reinforcement learning without parameter updates. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Theoretical memory scaling analysis, not deployment safety or robustness.
Provable Transformer Context Compression for In-Context Reinforcement Learning The paper analyzes the minimum context size required for transformer models to perform in-context reinforcement learning without parameter updates. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on theoretical bounds rather than benchmark evaluation.
On the Convergence of Semi-Gradient TD: A Mean Field Perspective The paper provides a theoretical analysis of stability and convergence for semi-gradient temporal-difference learning using mean-field neural networks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses mathematical stability in RL, not AI safety deployment.
On the Convergence of Semi-Gradient TD: A Mean Field Perspective The paper provides a theoretical analysis of stability and convergence for semi-gradient temporal-difference learning using mean-field neural networks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Mathematical proof paper rather than LLM benchmarking.
On the Convergence of Semi-Gradient TD: A Mean Field Perspective The paper provides a theoretical analysis of stability and convergence for semi-gradient temporal-difference learning using mean-field neural networks. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on RL convergence theory rather than efficient inference.
On the Convergence of Semi-Gradient TD: A Mean Field Perspective The paper provides a theoretical analysis of stability and convergence for semi-gradient temporal-difference learning using mean-field neural networks. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Theoretical analysis of reinforcement learning convergence, not agents.
MorphAct: Robot Policies Morphing with the Worlds They Act In The paper proposes a contextual parameterization framework that dynamically adjusts robot policy parameters as interaction contexts change. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses policy generalization rather than safety guardrails or alignment.
MorphAct: Robot Policies Morphing with the Worlds They Act In The paper proposes a contextual parameterization framework that dynamically adjusts robot policy parameters as interaction contexts change. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Focuses on general robot manipulation rather than industrial quality control.
MorphAct: Robot Policies Morphing with the Worlds They Act In The paper proposes a contextual parameterization framework that dynamically adjusts robot policy parameters as interaction contexts change. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes a robotics parameterization framework rather than a benchmark.
Memorizing Who, Knowing How: Benchmarking Personal and Procedural Memory for Persistent Assistants The paper introduces a benchmark to evaluate how personal assistants handle interdependent personal and procedural memories over long trajectories. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on memory maintenance benchmarks rather than safety risks.
SWE-Nexus: Automated Composition of Software Engineering Tasks from Test Execution The paper introduces an automated pipeline that builds repository-level software engineering benchmarks from test execution footprints. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on benchmark creation rather than safety or guardrails.
GeoCloak: An Agentic Framework for Semantics-Preserving Geolocation Privacy via Peripheral Clue Manipulation The paper introduces an agentic framework that removes peripheral visual clues to protect geolocation privacy in images without adding visible artifacts. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Introduces a defense framework rather than a general evaluation benchmark.
GeoCloak: An Agentic Framework for Semantics-Preserving Geolocation Privacy via Peripheral Clue Manipulation The paper introduces an agentic framework that removes peripheral visual clues to protect geolocation privacy in images without adding visible artifacts. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Focuses on geolocation privacy rather than industrial defect detection.
Beyond Correctness: Compute-Responsive Data Selection for Reasoning Model Post-Training The paper introduces a data selection strategy that prioritizes training examples where reasoning performance improves with additional inference compute. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Optimizes reasoning compute rather than addressing safety or jailbreaks.
Beyond Correctness: Compute-Responsive Data Selection for Reasoning Model Post-Training The paper introduces a data selection strategy that prioritizes training examples where reasoning performance improves with additional inference compute. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on post-training data selection, not autonomous agents.
Beyond Correctness: Compute-Responsive Data Selection for Reasoning Model Post-Training The paper introduces a data selection strategy that prioritizes training examples where reasoning performance improves with additional inference compute. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes data selection for training rather than evaluation benchmarks.
FoveaVLA: Foveation Actions Replace the Wrist Camera in VLAs The paper proposes a vision-language-action model that dynamically crops high-resolution regions to manipulate objects without using a wrist camera. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Evaluates general robotic manipulation rather than industrial visual inspection.
FoveaVLA: Foveation Actions Replace the Wrist Camera in VLAs The paper proposes a vision-language-action model that dynamically crops high-resolution regions to manipulate objects without using a wrist camera. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes a VLA model rather than benchmark evaluation frameworks.
FoveaVLA: Foveation Actions Replace the Wrist Camera in VLAs The paper proposes a vision-language-action model that dynamically crops high-resolution regions to manipulate objects without using a wrist camera. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on robotic manipulation policy, not safety or guardrails.
Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts The paper identifies an activation steering vector that controls whether language models rely on internal memory or retrieved external context. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Studies factual knowledge conflicts, not code generation or repair.
Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts The paper identifies an activation steering vector that controls whether language models rely on internal memory or retrieved external context. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Analyzes internal residual steering rather than agentic tool use.
Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts The paper identifies an activation steering vector that controls whether language models rely on internal memory or retrieved external context. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Exclusively investigates text representations in language models.
Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts The paper identifies an activation steering vector that controls whether language models rely on internal memory or retrieved external context. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on representation steering rather than benchmarking methodology.
Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts The paper identifies an activation steering vector that controls whether language models rely on internal memory or retrieved external context. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not explore model compression or efficient serving methods.
Trainable at Depth, Solvable at Small Size: What the Loss Surface of a Polynomial Activation Dictates for Designing, Initializing and Training Networks The paper investigates the loss surface of deep networks using a specific polynomial activation function without standard normalization layers. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Theoretical study of activations without focus on model compression.
Trainable at Depth, Solvable at Small Size: What the Loss Surface of a Polynomial Activation Dictates for Designing, Initializing and Training Networks The paper investigates the loss surface of deep networks using a specific polynomial activation function without standard normalization layers. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Studies optimization theory rather than AI safety or guardrails.
Trainable at Depth, Solvable at Small Size: What the Loss Surface of a Polynomial Activation Dictates for Designing, Initializing and Training Networks The paper investigates the loss surface of deep networks using a specific polynomial activation function without standard normalization layers. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Theoretical analysis of polynomial activations, not agentic workflows.
Trainable at Depth, Solvable at Small Size: What the Loss Surface of a Polynomial Activation Dictates for Designing, Initializing and Training Networks The paper investigates the loss surface of deep networks using a specific polynomial activation function without standard normalization layers. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on loss surfaces rather than LLM evaluation benchmarks.
Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlow The paper studies the impact of separating prediction horizons from state update intervals during sampling in generative flow models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address safety, security, or deployment reliability.
Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlow The paper studies the impact of separating prediction horizons from state update intervals during sampling in generative flow models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on sampling algorithms rather than evaluation benchmarks.
Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlow The paper studies the impact of separating prediction horizons from state update intervals during sampling in generative flow models. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Investigates diffusion sampling, not multi-step language agents.
Designing Textual Interfaces To Enhance Llm Reasoning On Audio-Visual Question Answering The paper compares three textual interface designs for optimizing prompts in multimodal audio-visual question answering systems. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on audio-visual reasoning, not software code generation.
Designing Textual Interfaces To Enhance Llm Reasoning On Audio-Visual Question Answering The paper compares three textual interface designs for optimizing prompts in multimodal audio-visual question answering systems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not study safety, guardrails, or jailbreak defenses.
Designing Textual Interfaces To Enhance Llm Reasoning On Audio-Visual Question Answering The paper compares three textual interface designs for optimizing prompts in multimodal audio-visual question answering systems. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not investigate model quantization, pruning, or efficient inference.
Designing Textual Interfaces To Enhance Llm Reasoning On Audio-Visual Question Answering The paper compares three textual interface designs for optimizing prompts in multimodal audio-visual question answering systems. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Studies prompt optimization rather than retrieval-augmented generation.
Designing Textual Interfaces To Enhance Llm Reasoning On Audio-Visual Question Answering The paper compares three textual interface designs for optimizing prompts in multimodal audio-visual question answering systems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Tests prompt optimization interfaces rather than benchmark methodologies.
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining The paper proposes a synthetic data pipeline that extracts atomic facts from source texts to guide factual pretraining for language models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Evaluates parametric pretraining knowledge rather than runtime retrieval augmentation.
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining The paper proposes a synthetic data pipeline that extracts atomic facts from source texts to guide factual pretraining for language models. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Studies pretraining data synthesis rather than autonomous multi-step agents.
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining The paper proposes a synthetic data pipeline that extracts atomic facts from source texts to guide factual pretraining for language models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Evaluates pretraining methods rather than proposing new LLM benchmarks.
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining The paper proposes a synthetic data pipeline that extracts atomic facts from source texts to guide factual pretraining for language models. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on general factual synthesis rather than code generation.
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining The paper proposes a synthetic data pipeline that extracts atomic facts from source texts to guide factual pretraining for language models. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on fact conditioning rather than model compression techniques.
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining The paper proposes a synthetic data pipeline that extracts atomic facts from source texts to guide factual pretraining for language models. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Addresses text pretraining rather than multimodal models.
EvalBuilder: Can LLMs Build Efficient Benchmarks? The paper presents a framework to evaluate whether language model agents can build lower-cost benchmark tests that match reference evaluations. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on benchmark generation rather than deployment safety or guardrails.
Task-State Governance: A Virtual Machine for Verifiable and Recoverable GUI Agent Execution The paper proposes an interface framework that tracks and controls application state changes made by graphical user interface agents. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents an execution system and contract framework, not LLM evaluation benchmarks.
From Spectral to Frobenius: When Muon Loses Its Spectral Implicit Bias The paper analyzes the mathematical implicit bias of practical Muon optimization compared to gradient descent on linearly separable classification problems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Theoretical analysis of optimization algorithms, not LLM benchmarking.
From Spectral to Frobenius: When Muon Loses Its Spectral Implicit Bias The paper analyzes the mathematical implicit bias of practical Muon optimization compared to gradient descent on linearly separable classification problems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Theoretical optimization paper, not AI safety or deployment reliability.
From Spectral to Frobenius: When Muon Loses Its Spectral Implicit Bias The paper analyzes the mathematical implicit bias of practical Muon optimization compared to gradient descent on linearly separable classification problems. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Studies theoretical implicit bias in optimizers, not LLM agents.
ACR-Net: Mitigating Semantic Dominance via Contrastive Acoustic-Semantic Decoupling The paper presents a network to decouple acoustic and semantic representations when speech tone conflicts with transcript text. Peer-reviewed	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Deals with speech emotion recognition, not AI deployment safety.
ACR-Net: Mitigating Semantic Dominance via Contrastive Acoustic-Semantic Decoupling The paper presents a network to decouple acoustic and semantic representations when speech tone conflicts with transcript text. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Studies acoustic-semantic feature decoupling, not LLM evaluation benchmarks.
ACR-Net: Mitigating Semantic Dominance via Contrastive Acoustic-Semantic Decoupling The paper presents a network to decouple acoustic and semantic representations when speech tone conflicts with transcript text. Peer-reviewed	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on speech emotion recognition, not multi-step LLM agents.

7,861–7,920 of 17,313← Previous Next →