Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,312In brief24More (on-topic)3,333Not relevant13,753Not assessed87Older351Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| KV4-RL: Unlocking Faster Asynchronous Rollouts for RLVR with INT4 KV Cache Quantization The paper introduces INT4 KV cache quantization with selective INT8 protection to accelerate asynchronous rollouts in reinforcement learning. Ranked #11Surprising resultPreprint7/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Introduces INT4 KV cache quantization to accelerate rollouts and reduce memory usage during inference. | |
| The Agent Knew: Internal States Forecast Unauthorized Actions and Stop Them, and the Lies That Follow The paper demonstrates that internal activation states can forecast and block unauthorized actions and deceptive reports by language model agents. Ranked #11Surprising resultPreprint8/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies agents that take unauthorized actions and attempt to misreport them. | |
| Can an Agent Learn to Predict Its Own Failure? Outcome-Grounded Confidence Calibration in LLM Agents The authors develop a reinforcement learning method to calibrate an agent's per-step confidence against the final outcome of an episode. Ranked #11New way to see a problemPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Provides reliability mechanisms for deciding when to trust an agent or abstain in deployment. | |
| 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. Ranked #11Surprising resultPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Analyzes LLM-as-a-judge consensus, rubric clarification, and alignment with human evaluations. | |
| Revisiting Generative vs. Dense Retrieval: A Static Value Prior Reverses the Gap Shows that incorporating static item value into dense retrieval outperforms generative retrieval in e-commerce recommendation settings. Ranked #12Surprising resultPreprint7/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06 | On-topic Compares generative versus dense retrieval mechanisms and reranking performance on retrieval datasets. | |
| EviScene: Beyond "Is It AI?" for Agentic Image Factuality Assessment Introduces a benchmark and multimodal agent that uses visual inspection and external retrieval to verify real-world image factuality. Ranked #12New testPreprint7/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Introduces a multimodal agent and benchmark combining visual inspection with external text retrieval. | |
| EdgeWise: Bringing Long-Horizon Agents to the Edge via Selective Cloud Delegation The paper proposes an edge-cloud collaboration framework that selectively delegates complex agent steps from local models to cloud models. Ranked #12New way to see a problemPreprint7/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Enables edge deployment of small language models by selectively offloading critical steps to reduce cost. | |
| Can an Agent Learn to Predict Its Own Failure? Outcome-Grounded Confidence Calibration in LLM Agents The authors develop a reinforcement learning method to calibrate an agent's per-step confidence against the final outcome of an episode. Ranked #12New way to see a problemPreprint8/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Calibrates per-step confidence in acting LLM agents to predict episode failure and guide handoffs. | |
| On the Risks of using LLM-Generated Tests for Regression Testing Investigates the risks of using LLM-generated regression tests, showing they can inadvertently preserve and enforce code faults. Ranked #12Surprising resultPreprint8/10 suggested | LLMs for code via arxiv · 12 Oct, 16:29 | On-topic Examines risks of using LLMs to generate regression tests for software repositories. | |
| What Visual Representations Do World Action Models Need? Evaluates visual representations and integration mechanisms for video-based world action models in robot manipulation tasks. Ranked #12Surprising resultPreprint6/10 suggested | Computer vision for industrial applications via arxiv · 12 Oct, 16:28 | On-topic Evaluates visual representations for world action models and manipulation tasks relevant to robotics perception. | |
| Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought Faithfulness Audits legal chain-of-thought faithfulness by substituting cited authorities and observing whether model verdicts causally change. Ranked #12Surprising resultPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Audits benchmark faithfulness and evaluation reliability using counterfactual authority perturbations. | |
| 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. Ranked #12Surprising resultPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Evaluates reliability and failure modes in automated LLM-based evaluation and supervision. | |
| Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling Develops small multimodal models that output structured choices using masked language modeling to reduce inference latency in agent tasks. Ranked #13New way to see a problemPreprint7/10 suggested | Small and efficient models via arxiv · 17 Oct, 06:17 | On-topic Develops small 0.8B-4B models optimized for low-latency decision-making inference. | |
| ExeWorldBench: Benchmarking Code-Based Executable 3D World Modeling The paper introduces a benchmark evaluating whether coding agents can construct interactive, code-based 3D simulations that capture underlying physical principles. Ranked #13New testPreprint7/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Benchmarks translating visual observations into code-based executable 3D environments. | |
| Retrieval Disagreement as a Signal: Diagnosing and Steering Multilingual Dense Embedders Analyzes retrieval divergence across languages in multilingual dense embedders and introduces disagreement-guided steering to align shared-index search results. Ranked #13Surprising resultPreprint7/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | On-topic Diagnoses and improves multilingual dense retrievers used in cross-lingual search and RAG. | |
| What Counts as the Same Task? Executable Contracts for Robustness Evaluation of Tool-Using Agents The paper proposes executable contracts to verify that adversarial benchmark edits preserve the core task definition when evaluating tool-using agents. Ranked #13New way to see a problemPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Analyzes flaws in agent robustness evaluation benchmarks and proposes executable contracts. | |
| CASTER: Formalizing Zero-Shot Manipulation from Synthetic Videos as Constrained Optimization Formulates zero-shot robot manipulation from synthetic video generation as a constrained trajectory optimization problem. Ranked #13New way to see a problemPreprint6/10 suggested | Computer vision for industrial applications via arxiv · 12 Oct, 16:28 | On-topic Develops robot manipulation techniques based on video models for physical robotics perception. | |
| Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought Faithfulness Audits legal chain-of-thought faithfulness by substituting cited authorities and observing whether model verdicts causally change. Ranked #13Surprising resultPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Tests whether model reasoning faithfully reflects cited authorities, exposing risks of unfaithful justifications. | |
| Co-Modifying Application Code and Deployment Infrastructure: A Causal Study of AI Agent Pull Requests Evaluates pull requests from software agents modifying both application code and deployment infrastructure in continuous integration pipelines. Ranked #13Surprising resultPreprint8/10 suggested | AI agents via arxiv · 12 Oct, 16:28 | On-topic Studies how software engineering agents handle multi-domain tasks and deployment infrastructure modifications. | |
| SWE-DeceptionBench: Benchmarking Deceptive Behaviors of LLMs Across the Software Development Lifecycle Introduces SWE-DeceptionBench, a benchmark measuring active and passive deceptive model behaviors across software development lifecycle tasks. Ranked #13New testPreprint8/10 suggested | LLMs for code via arxiv · 12 Oct, 16:28 | On-topic Evaluates LLM behaviors across the software engineering lifecycle including requirement feasibility and code tasks. | |
| Do Modifications Made by Coding Agents Make Subsequent Software Maintenance Harder? The paper evaluates whether initial code changes made by coding agents make subsequent software maintenance tasks harder than human changes. Ranked #14Surprising resultPreprint8/10 suggested | LLMs for code via arxiv · 15 Oct, 06:19 | On-topic Studies whether code generated by coding agents harms subsequent software maintenance. | |
| Elephant in the Fridge: Constant-Memory Frame Packing for Long Video Understanding Proposes a training-free frame packing method enabling vision-language models to process long videos with constant memory consumption. Ranked #14Surprising resultPreprint7/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Enables constant-memory efficient inference for long video vision-language models without fine-tuning. | |
| Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal Evidence Develops an agentic video reasoning framework that iteratively identifies missing information, gathers visual evidence, and grounds temporal clues. Ranked #14New way to see a problemPreprint7/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Develops an agentic grounded reasoning framework enabling multimodal video models to retrieve evidence and verify reasoning. | |
| Specification-Based Testing of Large Language Models The paper introduces a testing framework that turns documented model safety claims into specifications to find counterexamples via mutated prompts. Ranked #14Surprising resultPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a specification-driven testing framework to rigorously evaluate whether benchmarks test documented claims. | |
| What Counts as the Same Task? Executable Contracts for Robustness Evaluation of Tool-Using Agents The paper proposes executable contracts to verify that adversarial benchmark edits preserve the core task definition when evaluating tool-using agents. Ranked #14New way to see a problemPreprint8/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates the robustness of tool-using language model agents under adversarial changes. | |
| (When) Should a Software Factory be Lights-off? Benchmarking Code Review Agents on Bug Detection Introduces a benchmark pairing pull requests with human-verified bugs to evaluate code review agents on bug detection accuracy. Ranked #14New testPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:29 | On-topic Evaluates the reliability and bug detection capabilities of autonomous code review agents deployed in software engineering. | |
| Dual-Stream Memory for Vision-Language-Action Policies Presents a dual-stream memory architecture combining recurrent compression and frame retrieval for vision-language-action robotic policies. Ranked #14New way to see a problemPreprint5/10 suggested | Computer vision for industrial applications via arxiv · 12 Oct, 16:28 | On-topic Applies vision-language-action policies to robotic manipulation and physical execution tracking. | |
| Internalizer: Portable Context-to-Parameter Mapping for Very Large Language Models Presents a hypernetwork that maps document contexts directly into low-rank adapter weights for large language models. Ranked #14Surprising resultPreprint7/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28 | On-topic Maps external context documents directly into model parameters as an alternative to context-window RAG. | |
| Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Introduces a framework where a small model drafts tool-use trajectories and a larger model verifies them in single prefill passes. Ranked #15New way to see a problemPreprint7/10 suggested | Small and efficient models via arxiv · 17 Oct, 06:17 | On-topic Accelerates agent inference by delegating drafting to smaller models verified via parallel prefill. | |
| What Should a Software Team Spend on Agents? Simulation-Based Cost Optimization of Mixed AI-Human Teams The paper models mixed human-agent software teams to find cost-optimal spending allocations balancing human developer hours and automated coding agents. Ranked #15New way to see a problemPreprint8/10 suggested | LLMs for code via arxiv · 15 Oct, 15:07 | On-topic Investigates allocating budgets and roles between human developers and AI coding agents. | |
| VIEWCRAFT: Evolving What Multimodal Agents See and How They Read It This paper proposes a framework that jointly evolves visual evidence presentation and reading procedures to improve multimodal agent task performance. Ranked #15New way to see a problemPreprint7/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Evolves visual observation interfaces and reading procedures for multimodal agents. | |
| Measuring the Inference Energy Efficiency of Open-Weight LLMs across Benchmarks, Reasoning Modes, and Hardware The paper measures inference energy efficiency across 86 open-weight language models, showing large efficiency gaps among models of similar capability. Ranked #15Surprising resultPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Proposes methodology to measure energy efficiency alongside standard benchmark scores across models. | |
| Frustratingly Easy Agentic RAG via Portable Skills: A Case Study in Customer Support The paper presents an enterprise case study showing that portable skills and structured navigation layers simplify retrieval pipelines in customer support. Ranked #15Surprising resultPreprint8/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Presents an agentic workflow using tool use, instructions, and harnesses for enterprise customer support. | |
| What Counts as the Same Task? Executable Contracts for Robustness Evaluation of Tool-Using Agents The paper proposes executable contracts to verify that adversarial benchmark edits preserve the core task definition when evaluating tool-using agents. Ranked #15New way to see a problemPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Demonstrates how improper testing obscures actual agent vulnerabilities to hostile inputs. | |
| SOAR: Source-Aware Reinforcement Learning with Precise Credit Assignment for Search Agents Proposes a reinforcement learning framework that tracks external versus internal knowledge sources to reduce search agent fabrication. Ranked #15New way to see a problemPreprint7/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28 | On-topic Addresses search grounding and mitigating internal evidence fabrication using external retrieved evidence. | |
| 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. Ranked #15Surprising resultPreprint5/10 suggested | Computer vision for industrial applications via arxiv · 12 Oct, 16:28 | On-topic Addresses robotics manipulation and assembly skills for industrial benchmark tasks. | |
| Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling Develops small multimodal models that output structured choices using masked language modeling to reduce inference latency in agent tasks. Ranked #16New way to see a problemPreprint7/10 suggested | Multimodal models via arxiv · 17 Oct, 06:17 | On-topic Evaluates models making structured decisions from text and visual image inputs. | |
| BenchChecker: Assessing the Credibility of Bug-Fixing Benchmarks for LLMs This paper proposes a framework to detect training data contamination in bug-fixing benchmarks for language models. Ranked #16Surprising resultPeer-reviewed8/10 suggested | LLMs for code via arxiv · 15 Oct, 15:07 | On-topic Investigates data contamination detection in automatic program repair benchmarks for LLMs. | |
| The Silent Failures in Agentic Data Science This paper analyzes methodological failure modes in data science agents, such as data leakage and invalid evaluation, beyond simple code accuracy. Ranked #16Surprising resultPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces an extensible test suite evaluating methodological failures in agentic systems. | |
| The Silent Failures in Agentic Data Science This paper analyzes methodological failure modes in data science agents, such as data leakage and invalid evaluation, beyond simple code accuracy. Ranked #16Surprising resultPreprint8/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Analyzes how data science agents fail during end-to-end task formulation and execution. | |
| Specification-Based Testing of Large Language Models The paper introduces a testing framework that turns documented model safety claims into specifications to find counterexamples via mutated prompts. Ranked #16Surprising resultPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Provides a testing framework to verify safety behaviors, documented safety claims, and jailbreak counterexamples. | |
| Rethinking Projector Training in Multimodal LLMs The paper examines projector training in multimodal models and proposes a training-free initialization method using summary statistics instead of paired image-caption data. Ranked #16Surprising resultPreprint7/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Enables training-free projector initialization without gradient optimization or paired data, saving compute. | |
| StereoEngine: Scaling Stereo Bimanual Robot Data without Stereo Cameras Synthesizes stereo manipulation data from monocular robot demonstrations using estimated depth and video diffusion models. Ranked #16New way to see a problemPreprint5/10 suggested | Computer vision for industrial applications via arxiv · 12 Oct, 16:28 | On-topic Synthesizes stereo vision data for robotic bimanual manipulation. | |
| KV-Repurpose: A Testbed for Evaluating Non-Prefix KV-Cache Reuse Develops an open-source testbed to evaluate segment-wise key-value cache reuse beyond exact prompt prefixes during language model serving. Ranked #16Surprising resultPreprint7/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28 | On-topic Directly evaluates non-prefix KV-cache reuse tailored for repeated documents in RAG-style workloads. | |
| Elephant in the Fridge: Constant-Memory Frame Packing for Long Video Understanding Proposes a training-free frame packing method enabling vision-language models to process long videos with constant memory consumption. Ranked #17Surprising resultPreprint7/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Compresses arbitrarily long video sequences into visual tokens for vision-language models. | |
| Welcome to the Age of Agent Ads: Using On-Policy Attacks to Steer Agentic Decisions The paper investigates on-policy indirect manipulation attacks that steer autonomous agent purchasing and ranking decisions via web retrieval. Ranked #17Surprising resultPreprint8/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies attacks that manipulate web search and decision-making in tool-using agents. | |
| TurnCost: Turns, Not Observations: Cache-Aware Loop Engineering for Coding Agents Analyzes coding-agent trajectories to model execution costs under prompt caching, showing turn count matters far more than observation token length. Ranked #17Surprising resultPeer-reviewed8/10 suggested | LLMs for code via arxiv · 15 Oct, 06:17 | On-topic Investigates execution traces, loop engineering, and cost optimization of software engineering agents. | |
| History Replays Itself: Unleashing Long-Context Capability in Linear Attention The paper proposes a training-free episodic memory mechanism that extends linear attention language models to long-context tasks. Ranked #17New way to see a problemPreprint7/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Enables efficient long-context modeling in linear attention while dramatically reducing memory footprint. | |
| Do Factuality Methods Make LLMs More Factual? Reproducibility and Fair Comparison The authors reproduce and compare five language model factuality alignment methods on a standardized benchmark across identical base models and evaluation protocols. Ranked #17Surprising resultPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Provides reproducible evaluation revealing major benchmarking inconsistencies in factuality methods. | |
| Lost in Implicit Routing: How LLM Agents Fail to Load Relevant Skills at Scale The paper shows that large language model agents fail to retrieve relevant skills when large numbers of irrelevant skills populate the metadata library. Ranked #17Surprising resultPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Examines failure modes and reliability of LLM agents when selecting skills in production. | |
| VersaCamVLA: Camera-Configurable VLA Policies for Robotic Manipulation Proposes a camera-configurable framework for vision-language-action policies that handles variable camera counts and poses without fixed setups. Ranked #17New way to see a problemPreprint5/10 suggested | Computer vision for industrial applications via arxiv · 12 Oct, 16:28 | On-topic Presents camera-configurable robot manipulation policies relevant to industrial automation and robotics. | |
| Rehearse Everything, Remember Nothing: Attic-KV Rehearses What Will Be Read The paper shows that whole-context KV cache rehearsal harms compression under tight budgets, proposing targeted rehearsal to preserve answer-relevant entries. Ranked #17Surprising resultPreprint7/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28 | On-topic Optimizes KV cache retention for long-context retrieval and memory evaluation. | |
| SAFE-KV: Causal Suffix Repair for Behavior-Preserving Non-Prefix KV Cache Reuse in RAG The paper presents a repair method that re-encodes causal suffixes to enable safe non-prefix KV-cache reuse in retrieval-augmented generation. Ranked #18Surprising resultPreprint7/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | On-topic Addresses KV cache reuse specifically across reused documents in RAG pipelines. | |
| Cheap, Consistent, Calibrated, but Biased? Typed Decision Models as Judges The paper evaluates typed decision models against standard language model judges to measure how emotional tone and biases affect their evaluation accuracy. Ranked #18Surprising resultPreprint8/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Analyzes biases, calibration, and reliability in LLM-as-a-judge decision models. | |
| Receiver Decides Agent-to-Agent Negotiations Studies agent-to-agent negotiations over money, finding that the receiving agent's characteristics primarily dictate the final transfer amounts. Ranked #18Surprising resultPreprint8/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Analyzes multi-agent interactions and negotiation behavior in resource-exchange games. | |
| The Silent Failures in Agentic Data Science This paper analyzes methodological failure modes in data science agents, such as data leakage and invalid evaluation, beyond simple code accuracy. Ranked #18Surprising resultPreprint8/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Examines silent methodological failures like data leakage that affect reliable deployment. | |
| Executable Prefix Values for Locating Agent Failures The paper evaluates judge labels on agent steps by restarting failed trajectories from intermediate states to identify true failure points. Ranked #18Surprising resultPreprint8/10 suggested | LLMs for code via arxiv · 15 Oct, 06:17 | On-topic Evaluates failure localization on SWE-bench Verified coding agent trajectories. | |
| Virtual Head Attention Virtual Head Attention uses two linear operations to compress key-value cache overhead without increasing training compute or sacrificing output quality. Ranked #18New way to see a problemPreprint7/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Designs an efficient attention mechanism reducing KV cache overhead and compute during inference. | |
| A Minimal Optical-Flow Representation for Vision-Based Tactile Rotation Classification in Robotic Manipulation Across Gravity Domains The paper proposes a compact optical-flow representation to classify tactile object rotation across Earth, Mars, Moon, and orbital gravity domains. Ranked #18Surprising resultPreprint5/10 suggested | Computer vision for industrial applications via arxiv · 12 Oct, 16:28 | On-topic Investigates tactile vision representations for robotic manipulation and object slip/rotation detection. | |
| WOVEN: Weaving Visual World Modeling into Multimodal LLMs Presents a benchmark and dataset using generative video rollouts to train and evaluate visual transition reasoning in multimodal models. Ranked #18New testPreprint7/10 suggested | Multimodal models via arxiv · 12 Oct, 16:28 | On-topic Explicitly targets multimodal LLMs to enhance spatial, embodied, and temporal visual reasoning. |
61–120 of 3,333← PreviousNext →