Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,313In brief24More (on-topic)3,333Not relevant13,754Not assessed87Older352Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| PROBE: Learning to Audit Policy Compliance in Tool-Using LLM Agents The paper trains language model auditors to detect multi-turn policy violations during tool use by autonomous agents. Ranked #290New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Focuses on red-teaming and verifying policy safety constraints in deployed agent interactions. | |
| Vision-Predictor-Action: Towards Efficient Embodied AI via Neuro-Symbolic JEPA Predictors The paper proposes a robot control framework using predictive architectures and flow matching to generate action chunks without autoregressive decoding. Ranked #290New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Proposes a lightweight non-autoregressive architecture under 20M parameters for fast, efficient embodied AI control. | |
| D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Ranked #290New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Develops full-duplex spoken language models processing speech and text. | |
| Knowledge Graphs as Files: An Agent Harness Where Tool Creation and Tool Use Co-Evolve This paper proposes a system where language model agents write code to create and refine tools directly over raw knowledge-graph files. Ranked #290New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Develops an agent harness where agents write Python tools to navigate knowledge graphs. | |
| Mind the Spatial Gap: A Benchmark for Geospatial Reasoning and Constrained Spatial Editing Presents a benchmark using real urban data to evaluate multimodal language models on spatial reasoning and constrained editing. Ranked #290New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces GeoSpatialBench, a comprehensive benchmark evaluating multimodal spatial reasoning. | |
| Prism-1: Position- and Rival-Independent Scoring for System One Models The paper evaluates single-pass decision models against rational choice axioms and identifies ordering and rival-option sensitivity failures. Ranked #291Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Investigates decision-making reliability, manipulation vulnerabilities, and consistency for deployed systems. | |
| Error Correction in Trained TTT Layers This paper analyzes prediction-derived error updates and test-time gradient objectives in trained test-time-training layers. Ranked #291Surprising resultPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Analyzes fast-weight test-time training layers designed for efficient sequence processing. | |
| From Agents to Edges: Paired Optimization for Multi-Agent Collaboration The paper introduces an optimization method that jointly refines communication instructions between communicating agents to improve multi-agent collaboration. Ranked #291New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Optimizes communication edges and instruction alignment across multi-agent collaborative workflows. | |
| REEL: Learning to Retrieve, Evaluate, and Evolve Visual Memory for Any Long-Horizon Multi-Shot Video Generation Introduces a general visual memory planner that retrieves, creates, and updates visual evidence across shots for long-horizon video generation. Ranked #291New testPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Focuses on visual memory planning and multi-shot video generation across multimodal inputs. | |
| DRYLAB-Bench: Benchmarking the Biosecurity Risk of Agentic LLM Mutation Design Introduces a benchmark evaluating the risks of LLM agents designing dangerous biological mutations when augmented with biodesign tools. Ranked #291New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces a comprehensive benchmark to evaluate biosecurity risks in agentic biodesign settings. | |
| FAOPs: Normal Forms and an Algebra for AI Agent Workflows The paper provides a formal algebraic framework and normal forms to structure and audit enterprise AI agent workflows. Ranked #292New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Enforces least-privilege tool scoping, auditability, and safety boundaries for enterprise agent deployments. | |
| On the Depth Transfer of Neural Optimizers via L-Smoothness The paper analyzes depth transfer of hyperparameters in neural networks using an L-smoothness theoretical framework. Ranked #292New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Provides theoretical methods for scaling neural network depth and hyperparameter transfer efficiently. | |
| The Last Mile of Coding Agents for Long-Document Tasks: An Empirical Study on Clinical Trial Protocol Digitization Introduces a benchmark evaluating how reliably software coding agents extract and digitize complex unstructured clinical trial protocol documents. Ranked #292New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies coding agents executing multi-step workflows, tool calls, and document structuring. | |
| Across Viewpoints and Time: Trajectory-Generalizable Physical Attacks on World Models and Video Understanding Models The paper develops an attack framework to generate physical adversarial textures that deceive video models across varying trajectories and viewpoints. Ranked #292New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Studies cross-viewpoint adversarial robustness for dynamic video understanding and world models. | |
| CoBe: A Benchmark For Conversational Counterfactual Text Editing Presents a conversational benchmark testing whether models can edit text under counterfactual changes while preserving unaffected factual details. Ranked #292New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces CoBe, a novel benchmark evaluating zero-shot counterfactual text editing capabilities. | |
| Momentum, Commentary Tone, and Earnings Surprise in LLM Financial Forecasts The paper tests how language models weigh price momentum, commentary tone, and earnings surprises when forecasting financial outcomes. Ranked #293Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Assesses model reliability and misleading non-informational biases in high-stakes financial forecasting settings. | |
| Outcome-Routed Distillation and Reinforcement Learning for Diffusion Language Models The paper presents an outcome-routed distillation and reinforcement learning method tailored for training mixture-of-experts diffusion language models. Ranked #293New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Introduces outcome-routed distillation and reinforcement learning to improve diffusion language model efficiency. | |
| Flow-Support-Weighted Proposals for Test-Time Verified VLA Policies Introduces a training-time sampling method to prevent action proposal collapse in vision-language-action models tested with verifiers. Ranked #293New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Investigates action proposal scaling and flow decoders in vision-language-action models. | |
| ToolHACE: A Dataset for Detecting Hallucinations in Tool-Based LLM Responses This paper introduces a benchmark dataset to evaluate hallucination detection in language model responses after external tool execution. Ranked #293New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies hallucination in tool-augmented LLM responses during multi-step tool use. | |
| Predicting Alignment Generalization with Value Representations Evaluates whether internal model representations can predict how fine-tuning on one value generalizes across other held-out values. Ranked #293New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Establishes a new evaluation task and benchmark for predicting model generalization across alignment targets. | |
| Retrieved but Not Read: Understanding Why Search-Augmented LLMs Fail to Use Available Evidence The paper analyzes why search-augmented language models produce outdated answers even when relevant evidence is retrieved. Ranked #294Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Examines unfaithful citations and failure to ground claims in evidence. | |
| Speculative Sampling in Interactive Video Generation This paper proposes Speculative Sampling to combine large bidirectional sampling windows with low-latency interaction in interactive video generation. Ranked #294Surprising resultPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Addresses sampling and latency for interactive text/conditioned video generation. | |
| One Block Ahead: Promoting Sampler State in Diffusion Language Models The paper proposes a decoding algorithm that denoises consecutive blocks concurrently by promoting sampler states in diffusion language models. Ranked #294Surprising resultPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Improves diffusion language model serving efficiency and throughput via concurrent block decoding. | |
| Learn from Move: the Next Step for GUI Agents This paper introduces an environment benchmark and exploration framework to help graphical user interface agents learn unfamiliar interaction dynamics. Ranked #294New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Introduces a benchmark and training method for multi-step GUI exploration agents. | |
| ReSI: Recursive Safety Improvement toward Resistant and Resilient AI Presents ReSI, a recursive safety framework using cyclical evaluation and updates to improve AI resilience against emergent threats. Ranked #294New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Develops an automated red-teaming and safety evaluation framework to benchmark model updates. | |
| Human Interventions as Labels: Localizing, Steering, and (Sometimes) Editing the Computations Behind LLM Agent Failures The paper uses human interventions to localize and steer computations behind agent failure to decide between acting and clarifying. Ranked #295New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Improves agent reliability and oversight by diagnosing internal failure causes. | |
| Rethinking Routing Exploration in Modern Mixture-of-Experts The authors propose MirrorMoE to reconcile an exploration dilemma between fine-grained expert segmentation and shared experts in mixture-of-experts models. Ranked #295New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 17 Oct, 06:17 | On-topic Investigates expert routing efficiency and architecture design in Mixture-of-Experts models. | |
| What Matters for World Modeling in Vision-Language-Action Models? Evaluates four world-model integration paradigms to identify what improves vision-language-action policies in robotic tasks. Ranked #295New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Systematically analyzes world model integrations within Vision-Language-Action multimodal architectures. | |
| CivWorldBench: Evaluating Language World Models in Structured Multi-Agent Sandbox Worlds This paper introduces a multi-agent sandbox benchmark to evaluate language models on state inference, forecasting, and intervention prediction. Ranked #295New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Evaluates language models acting in multi-agent strategy game sandboxes and making decisions. | |
| InterviewPlayground: A Validated Simulation Environment for Evaluating AI Interviewers Presents a simulated participant environment to evaluate multi-turn AI interviewers across market research and social science scenarios. Ranked #295New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces a simulated environment and report card benchmark specifically to evaluate AI interviewer performance. | |
| DialFork: Going Beyond Factuality to Evaluate Procedural Hallucination in Conversational Agents The authors present a benchmark to evaluate conversational agents that follow correct factuality but violate procedural task steps. Ranked #296New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Evaluates conversational procedural hallucination, off-track dialogue detection, and real-time guardrails. | |
| SaTPO: Text-to-Image Safety Alignment via Transition Preference Optimization The paper proposes a step-wise alignment method to suppress harmful content in text-to-image models while maintaining generation semantics. Ranked #296New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Studies safety alignment and preference optimization in multimodal text-to-image diffusion models. | |
| WUSPruner: Second-Moment-Optimal Transforms for Structured Pruning This paper presents a transform-based framework for structured pruning in language models that optimizes basis selection using activation and weight statistics. Ranked #296New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Presents a structured pruning framework to make LLMs more efficient to run. | |
| RIVO: Lightweight Adaptive Rerouting for Deep Research Agents RIVO is a lightweight plug-in framework that adaptively reroutes research agent trajectories to prevent premature stopping or wasted steps. Ranked #296New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Focuses on adaptive trajectory rerouting for multi-step research agents. | |
| OccludeBench: Can VLMs Find What's Missing? The paper presents a benchmark evaluating how effectively vision-language models recover fully occluded objects using remaining visual cues. Ranked #296New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces OccludeBench, a curated benchmark evaluating visual reasoning on occluded objects. | |
| Following Delegation Traces to Study Human Oversight in Interaction with Multimodal AI Investigates how user trust influences delegation and verification choices during human interaction with multimodal video question-answering systems. Ranked #297New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Analyzes human oversight and operational trust policies for reliable AI deployment. | |
| StateCraft: Augmenting Video World Models with Abstract-State Reasoning This paper pairs language models with video diffusion to maintain persistent abstract states across dynamic video world interactions. Ranked #297New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Combines language models with video diffusion to model dynamic visual environments. | |
| Re-executing Coding-Agent Evaluations: Unchanged Decisions and Unresolved Policy Risk The paper re-evaluates coding-agent benchmark results to measure sensitivity to test execution environments and grading configurations. Ranked #297Surprising resultPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Audits evaluation reproducibility of autonomous coding agents. | |
| A Few More Tokens Ahead: Reinforcing Parallelism in Diffusion LLMs RePara trains diffusion language models using group-relative reinforcement learning on short token blocks to improve math generation speed and accuracy. Ranked #297New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Improves parallel generation and reduces inference and training costs in diffusion LLMs. | |
| Not All Correct Fixes Run Alike: An Empirical Study of Patch Performance in Repository-Level Issue Resolution The paper analyzes the runtime performance differences of test-passing issue resolution patches generated by language models on SWE-bench. Ranked #297New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Investigates repurposing FAIL_TO_PASS tests as a benchmark evaluation metric for patch performance. | |
| In the Guise of Correction: How Framing and Cultural Adaptation Erode LLM Safety on Real-World Criminal Cases Presents a benchmark based on real criminal cases showing how correction framings and cultural adaptations bypass safety guardrails. Ranked #298Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Measures LLM jailbreak vulnerabilities and unsafe compliance risks in criminal scenarios. | |
| VideoHarness-SRSI: Structured Recursive Self-Improvement for Video Understanding Harness The paper examines failure modes in recursively self-improving execution harnesses for agentic video understanding models. Ranked #298New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Improves vision-language model harnesses for multimodal video understanding tasks. | |
| SciCurateBench: Curating Experimental Data from Heterogeneous Scientific Literature with Verbal-Action Agents The paper presents SciCurateBench to evaluate language models on verifying complex experimental conditions from unprocessed scientific literature. Ranked #298New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Evaluates and develops multi-agent verbal-action systems for scientific curation. | |
| TabGift: A Context-Efficient Foundation Model for Tabular Data Generation The paper proposes a context-efficient foundation model for generating realistic tabular data from limited context observations. Ranked #298New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Focuses on context-efficient tabular generation from few observations and compact latent representations. | |
| Personalization Without Stereotyping: Do Speech LLMs Know When to Turn a Deaf Ear? The paper introduces a benchmark to measure demographic robustness, stereotyping, and intentional personalization in speech language models. Ranked #298New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces BASIS-Bench, a benchmark measuring demographic robustness, stereotypical bias, and personalization. | |
| Poison Valleys: Diagnosing Path-Dependent Failure in LLM Code Repair Analyzes how misconceptions in early planning stages cause cascading failures during automated code repair by language model agents. Ranked #299New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Analyzes internal failure propagation and reliability in automated code generation. | |
| Frame Evidence:A Benchmark for Reference Frames and Visual Evidence in Multi-Image Spatial Reasoning The paper introduces a spatial reasoning benchmark to evaluate how models handle varying reference frames and altered visual evidence. Ranked #299New testPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Evaluates multi-image visual-language spatial reasoning and grounding. | |
| OpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native Tools The paper presents OpenMobile-2, combining open mobile environments, trajectory datasets, and hybrid GUI-tool actions for mobile agents. Ranked #299New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Develops mobile agents that plan, interleave GUI actions, and call app-native tools. | |
| Before Attention Sinks Form: Hidden-State Geometry Reveals and Modulates Their Formation The paper analyzes pre-sink hidden-state geometry to detect and modulate attention sink formation in language models. Ranked #299Surprising resultPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Analyzes attention sink formation mechanisms that impact long-context streaming inference efficiency. | |
| What Output-Only Review Cannot Verify: Study Contracts for Research Agents Proposes study contracts to register execution obligations and verify research agent outputs against approved protocol parameters. Ranked #299New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Focuses on diagnostic verification and defect-checking methodology for automated research outputs. | |
| FlashAttention with Statistics: Low-Overhead Attention Features for Compact Hallucination Detectors Extracts attention statistics directly during fast attention computation to build lightweight hallucination detectors for language models. Ranked #300Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Addresses hallucination detection and safety guardrails for deployed LLMs with low overhead. | |
| AdaRSI: On-Demand Recursive Self-Improvement for Embodied Agents This paper introduces AdaRSI, an on-demand recursive self-improvement framework for embodied agents that isolates and fine-tunes only persistent sub-step failures. Ranked #300New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Improves vision-language-action models for multimodal perception and robotic action execution. | |
| 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. Ranked #300Surprising resultPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Analyzes the behavior and implementation mechanics of the efficient Muon optimizer. | |
| Evidence-First Agentic Verification Introduces Sauron, an evidence-first agentic verifier that checks both environment state and execution traces against original sources to validate task rubrics. Ranked #300New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Develops Sauron, an evidence-first agentic verifier to inspect agent execution traces. | |
| MultiWorldBench: Do Independently Controlled Views Describe One Shared World? Introduces a Minecraft diagnostic benchmark testing whether multiplayer world models maintain consistent and persistent representations across independently controlled views. Ranked #300New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces a diagnostic benchmark to test consistency across views in world models. | |
| Probe-and-Resume: When and How to Evaluate Agent Execution Evaluates test-time search in computer-use agents, finding critics perform better when invoked only at predicted final states. Ranked #301New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Investigates critic inspection and state probing to reliably assess agent execution and prevent failures. | |
| BluePixel: Toward Native 4K Agentic Image Editing Combines an adapted editing model with an agent harness to perform instruction-driven four-kilopixel image editing. Ranked #301New way to see a problemPreprint5/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Develops an agentic instruction-driven native 4K image editing vision-language system. | |
| How Inner Optimizers Shape Replica Scaling Studies how the choice of local optimizers affects model performance degradation when scaling distributed training replicas under fixed budgets. Ranked #301New way to see a problemPreprint4/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Investigates local optimizer designs to scale distributed training efficiently across constrained compute clusters. | |
| Keep Your Receipts: Trustworthy Cleanup of Agent-Built Knowledge Graphs Presents PolyGraph, a method storing provenance receipts during knowledge graph construction to enable precise updates when source documents are retracted. Ranked #301New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Presents a framework for maintaining and cleaning knowledge graphs built by agents. | |
| GeopoliticalQA: Benchmarking Retrieval and Generation over Intelligence Reports Introduces a benchmark of 5,525 QA pairs from intelligence reports to evaluate RAG document- and chunk-level retrieval. Ranked #301New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28 | On-topic Introduces GeopoliticalQA to systematically benchmark retrieval systems and LLM answer generation. |
1,741–1,800 of 3,333← PreviousNext →