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
YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation The paper introduces a framework combining 3D Gaussian Splatting and physical simulation to evaluate robot policies using handheld fisheye cameras. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Addresses simulation fidelity for robot policies rather than AI safety guardrails.
YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation The paper introduces a framework combining 3D Gaussian Splatting and physical simulation to evaluate robot policies using handheld fisheye cameras. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on robotic policy simulation evaluation, not language model benchmarks.
YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation The paper introduces a framework combining 3D Gaussian Splatting and physical simulation to evaluate robot policies using handheld fisheye cameras. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Evaluates robotic manipulation policies in simulation, not industrial inspection.
YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation The paper introduces a framework combining 3D Gaussian Splatting and physical simulation to evaluate robot policies using handheld fisheye cameras. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on real-to-sim 3DGS scene reconstruction for robotic evaluation, not LLM agents.
Enhanced Detective: A Dual-Stage Proactive Defense Against Internal Poisoning Attacks in Financial Fraud Detection Using Federated Learning The paper introduces a dual-stage robust aggregation defense against data and model poisoning attacks in federated financial fraud detection. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Studies poisoning defenses in FL, not LLM evaluation benchmarks.
Enhanced Detective: A Dual-Stage Proactive Defense Against Internal Poisoning Attacks in Financial Fraud Detection Using Federated Learning The paper introduces a dual-stage robust aggregation defense against data and model poisoning attacks in federated financial fraud detection. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on federated learning fraud models, not LLM agents.
Box-WAM: Box Trajectories as Compact State Representations for World Action Models The paper proposes predicting 3D bounding box trajectories instead of dense video frames as compact representations for world action models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on robotic state representation efficiency, not safety guardrails.
Box-WAM: Box Trajectories as Compact State Representations for World Action Models The paper proposes predicting 3D bounding box trajectories instead of dense video frames as compact representations for world action models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a policy modeling representation, not an LLM benchmark.
Box-WAM: Box Trajectories as Compact State Representations for World Action Models The paper proposes predicting 3D bounding box trajectories instead of dense video frames as compact representations for world action models. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on robotic manipulation policies, not industrial defect detection.
Towards Green AI: Evaluating Energy Efficiency for Quantized LLM Inference The paper evaluates the energy efficiency, carbon intensity, and accuracy of ten post-training quantization methods across various language models and GPUs. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on hardware energy profiling rather than LLM benchmark design.
Towards Green AI: Evaluating Energy Efficiency for Quantized LLM Inference The paper evaluates the energy efficiency, carbon intensity, and accuracy of ten post-training quantization methods across various language models and GPUs. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Studies energy and carbon metrics rather than safety, hallucinations, or robustness.
Towards Green AI: Evaluating Energy Efficiency for Quantized LLM Inference The paper evaluates the energy efficiency, carbon intensity, and accuracy of ten post-training quantization methods across various language models and GPUs. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Investigates quantization energy efficiency, not LLM autonomous agents.
TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning Latency The paper proposes a vision-language-action model where action execution avoids inference latency by decoupling from explicit reasoning output at runtime. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on reducing reasoning latency in robotics, not safety guardrails.
TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning Latency The paper proposes a vision-language-action model where action execution avoids inference latency by decoupling from explicit reasoning output at runtime. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Evaluates robotic manipulation and autonomous driving, not industrial visual inspection.
TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning Latency The paper proposes a vision-language-action model where action execution avoids inference latency by decoupling from explicit reasoning output at runtime. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Evaluates a specific model architecture, not benchmarking LLMs.
Convex-Analytic View of Query Selection for Safety Verification of Markov Decision Processes The paper presents a convex-analytic framework to select query directions for safety verification in Markov decision processes. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies formal verification of MDPs, not LLM autonomous agents.
Convex-Analytic View of Query Selection for Safety Verification of Markov Decision Processes The paper presents a convex-analytic framework to select query directions for safety verification in Markov decision processes. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on formal MDP verification rather than LLM safety or guardrails.
Convex-Analytic View of Query Selection for Safety Verification of Markov Decision Processes The paper presents a convex-analytic framework to select query directions for safety verification in Markov decision processes. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on mathematical verification algorithms, not LLM evaluation benchmarks.
Many Steps into One: Reasoning Contraction and the Limits of Safe Memory Reuse for LLM Agents The paper analyzes the failure modes of agent memory reuse and evaluates certified memory update strategies when facts change. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on agent reasoning contraction and skill caches rather than RAG.
Many Steps into One: Reasoning Contraction and the Limits of Safe Memory Reuse for LLM Agents The paper analyzes the failure modes of agent memory reuse and evaluates certified memory update strategies when facts change. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Presents theoretical and empirical bounds on memory safety, not an LLM benchmark.
Many Steps into One: Reasoning Contraction and the Limits of Safe Memory Reuse for LLM Agents The paper analyzes the failure modes of agent memory reuse and evaluates certified memory update strategies when facts change. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Does not involve multimodal reasoning or vision-language models.
Many Steps into One: Reasoning Contraction and the Limits of Safe Memory Reuse for LLM Agents The paper analyzes the failure modes of agent memory reuse and evaluates certified memory update strategies when facts change. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Uses code verifiers as a test case, not developing code LLMs.
Many Steps into One: Reasoning Contraction and the Limits of Safe Memory Reuse for LLM Agents The paper analyzes the failure modes of agent memory reuse and evaluates certified memory update strategies when facts change. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Focuses on reasoning reuse safety, not model compression or serving.
Reference Wording Shapes Retrieval of Robot Task Instructions The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on robotic instruction retrieval, not industrial quality control vision.
Reference Wording Shapes Retrieval of Robot Task Instructions The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not involve coding assistants, program repair, or code generation.
Reference Wording Shapes Retrieval of Robot Task Instructions The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Evaluates lexical ranker wording sensitivity, not general LLM benchmarks.
Reference Wording Shapes Retrieval of Robot Task Instructions The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Studies lexical retrieval sensitivity, not model compression or efficiency.
Reference Wording Shapes Retrieval of Robot Task Instructions The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Focuses on text-based lexical retrieval of robot task descriptions.
Reference Wording Shapes Retrieval of Robot Task Instructions The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Does not address deployment safety, jailbreaking, or red teaming.
Reference Wording Shapes Retrieval of Robot Task Instructions The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Examines instruction wording retrieval for robots, not LLM agent systems.
FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching The paper presents an efficient video super-resolution method using hybrid attention and reward-yield distribution matching distillation. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on general video super-resolution, not industrial defect detection.
FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching The paper presents an efficient video super-resolution method using hybrid attention and reward-yield distribution matching distillation. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on super-resolution quality and speed rather than safety guardrails.
FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching The paper presents an efficient video super-resolution method using hybrid attention and reward-yield distribution matching distillation. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Develops video super-resolution architectures, not LLM planning agents.
FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching The paper presents an efficient video super-resolution method using hybrid attention and reward-yield distribution matching distillation. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a video generation method, not an LLM benchmark.
Benchmarking Self-Supervised 3D Medical Vision Foundation Models and Transferability Estimation The paper benchmarks twenty-three self-supervised 3D medical vision foundation models across segmentation, classification, and regression tasks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on benchmark transferability across medical tasks, not safety risks.
Benchmarking Self-Supervised 3D Medical Vision Foundation Models and Transferability Estimation The paper benchmarks twenty-three self-supervised 3D medical vision foundation models across segmentation, classification, and regression tasks. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on medical vision models rather than industrial defect inspection.
Benchmarking Self-Supervised 3D Medical Vision Foundation Models and Transferability Estimation The paper benchmarks twenty-three self-supervised 3D medical vision foundation models across segmentation, classification, and regression tasks. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Evaluates 3D medical vision encoders, not multimodal vision-language models.
Benchmarking Self-Supervised 3D Medical Vision Foundation Models and Transferability Estimation The paper benchmarks twenty-three self-supervised 3D medical vision foundation models across segmentation, classification, and regression tasks. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Benchmarks 3D medical vision models, not language model agents.
Dissociating Answerability from Extraction in LLM Question Answering over Radiology Reports The paper introduces an evaluation method that separates question answerability detection from answer span extraction in radiology report question answering. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Evaluates QA extraction over existing reports without studying retrieval mechanisms.
Dissociating Answerability from Extraction in LLM Question Answering over Radiology Reports The paper introduces an evaluation method that separates question answerability detection from answer span extraction in radiology report question answering. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on medical report QA, not software engineering or code generation.
Dissociating Answerability from Extraction in LLM Question Answering over Radiology Reports The paper introduces an evaluation method that separates question answerability detection from answer span extraction in radiology report question answering. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Evaluates text QA over radiology reports, not multimodal models.
Dissociating Answerability from Extraction in LLM Question Answering over Radiology Reports The paper introduces an evaluation method that separates question answerability detection from answer span extraction in radiology report question answering. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies question answering evaluation, not multi-step autonomous LLM agents.
Dissociating Answerability from Extraction in LLM Question Answering over Radiology Reports The paper introduces an evaluation method that separates question answerability detection from answer span extraction in radiology report question answering. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Evaluates existing models rather than efficient architectures or compression.
Chasing Label Shifters: A Change-Aware Framework for Dynamic Graph Node Classification The paper analyzes why dynamic graph neural networks fail on evolving nodes and introduces a change-aware classification framework. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Addresses dynamic graph neural networks, not language model agents.
Chasing Label Shifters: A Change-Aware Framework for Dynamic Graph Node Classification The paper analyzes why dynamic graph neural networks fail on evolving nodes and introduces a change-aware classification framework. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Does not evaluate AI safety, hallucinations, jailbreaks, or guardrails.
Chasing Label Shifters: A Change-Aware Framework for Dynamic Graph Node Classification The paper analyzes why dynamic graph neural networks fail on evolving nodes and introduces a change-aware classification framework. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Focuses on label shifts in GNNs rather than model compression or efficiency.
Chasing Label Shifters: A Change-Aware Framework for Dynamic Graph Node Classification The paper analyzes why dynamic graph neural networks fail on evolving nodes and introduces a change-aware classification framework. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Investigates dynamic graph node classification rather than LLM evaluation benchmarks.
RAN-JEPA: An Action-Conditioned World Model for Energy-efficient Mobile Network Control The paper introduces an action-conditioned world model that separates observation representation learning from intervention learning for mobile network energy control. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Develops a world model for mobile network control, not language model agents.
RAN-JEPA: An Action-Conditioned World Model for Energy-efficient Mobile Network Control The paper introduces an action-conditioned world model that separates observation representation learning from intervention learning for mobile network energy control. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on energy-efficient mobile network control rather than AI safety.
RAN-JEPA: An Action-Conditioned World Model for Energy-efficient Mobile Network Control The paper introduces an action-conditioned world model that separates observation representation learning from intervention learning for mobile network energy control. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Evaluates a domain-specific network model, not general LLM benchmarks.
Taming Softmax at the Limits of Muon: A Unified Approach through Hilbert Geometry The paper applies Hilbert geometry to constrain activation shifts in softmax layers for stable optimizer updates in language model heads. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Investigates mathematical optimization metrics rather than deployment safety or robustness.
Taming Softmax at the Limits of Muon: A Unified Approach through Hilbert Geometry The paper applies Hilbert geometry to constrain activation shifts in softmax layers for stable optimizer updates in language model heads. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Presents an optimization formulation, not an evaluation method or benchmark.
Taming Softmax at the Limits of Muon: A Unified Approach through Hilbert Geometry The paper applies Hilbert geometry to constrain activation shifts in softmax layers for stable optimizer updates in language model heads. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on optimization mathematics for neural network training, not AI agents.
Revisiting the 16x Pretraining Compute Gap between Diffusion and Autoregressive Language Models The paper re-evaluates the pretraining compute scaling laws between diffusion and autoregressive language models using modern diffusion training designs. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies diffusion language model pretraining compute, not multi-step planning or agents.
Revisiting the 16x Pretraining Compute Gap between Diffusion and Autoregressive Language Models The paper re-evaluates the pretraining compute scaling laws between diffusion and autoregressive language models using modern diffusion training designs. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on pretraining compute scaling laws, not general LLM evaluation benchmarks.
Revisiting the 16x Pretraining Compute Gap between Diffusion and Autoregressive Language Models The paper re-evaluates the pretraining compute scaling laws between diffusion and autoregressive language models using modern diffusion training designs. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Does not address safety, guardrails, hallucinations, or deployment reliability.
Tanawpamana: Benchmarking Vision-Language Models on Philippine Cultural Heritage Sites The paper introduces a dataset and benchmark to evaluate vision-language models on recognizing and answering questions about Philippine cultural heritage sites. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Introduces an evaluation benchmark, not model compression or efficiency.
Tanawpamana: Benchmarking Vision-Language Models on Philippine Cultural Heritage Sites The paper introduces a dataset and benchmark to evaluate vision-language models on recognizing and answering questions about Philippine cultural heritage sites. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Evaluates cultural heritage site recognition, not software engineering.
Tanawpamana: Benchmarking Vision-Language Models on Philippine Cultural Heritage Sites The paper introduces a dataset and benchmark to evaluate vision-language models on recognizing and answering questions about Philippine cultural heritage sites. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Evaluates static vision-language perception, not autonomous agentic systems.
Tanawpamana: Benchmarking Vision-Language Models on Philippine Cultural Heritage Sites The paper introduces a dataset and benchmark to evaluate vision-language models on recognizing and answering questions about Philippine cultural heritage sites. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on image classification and knowledge, not RAG systems.

13,441–13,500 of 17,313← Previous Next →