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 |
|---|---|---|---|
| Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents The paper analyzes prefix replay curricula and handoff rules under reinforcement learning to improve long-horizon task completion for language model agents. Ranked #350New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Investigates curriculum strategies for training long-horizon tool-calling LLM agents. | |
| Repairing Speech Access After Text Updates with Frozen Affine Fields This paper introduces AFFIRM, using affine fields on frozen speech models to repair spoken access to recently updated text knowledge. Ranked #350New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Repairs speech modality access to knowledge updated via text in multimodal models. | |
| The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language? This paper investigates output diversity and text homogenization across twenty-three open-weight language models and thirteen languages. Ranked #350Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Examines model collapse and output homogenization risks in multilingual LLM generation. | |
| CodeWalk: Generating Coding Benchmarks by Walking a Problem Graph The paper presents CodeWalk, a method to create challenging coding benchmarks at low cost by programmatically composing verified atomic problems. Ranked #350New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a compositional method for constructing challenging coding evaluation benchmarks. | |
| COFFEQA: Counterfactual-Guided Few-Shot Distillation for LLM Question Answering Introduces a counterfactual-guided few-shot distillation framework to train smaller question-answering language models using limited training samples. Ranked #350Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Proposes a sample-efficient few-shot distillation method to train smaller student models. | |
| OS-Helix: Unifying GUI Interaction and Code Execution in Native Computer-Using Agents The paper presents an end-to-end trained multimodal agent that combines visual graphical interface actions and executable code execution. Ranked #351New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 17 Oct, 06:17 | On-topic Develops an autonomous agent unifying GUI navigation and code execution for desktop tasks. | |
| DynThinker: Unifying Static 3D and Dynamic 4D Spatial Reasoning This paper presents DynThinker, a framework using parameter averaging and spatial-dynamic fusion to balance static 3D and dynamic 4D reasoning. Ranked #351New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Integrates spatial and dynamic priors into multimodal vision-language representations. | |
| Intent-Triggered Anchoring for Multi-Token Format Units in Parallel Decoding of Diffusion Language Models This paper introduces a training-free decoding rule that anchors multi-token format units to ensure parseable parallel generation in diffusion language models. Ranked #351Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Mitigates parsing failures and ungrounded hallucinations caused by parallel token commitment. | |
| UDHRBench: Benchmarking Human Rights in Large Language Models Evaluates commercial language models against Universal Declaration of Human Rights principles, revealing misalignments driven by efficiency arguments. Ranked #351New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces UDHRBench and an LLM-as-a-judge framework to evaluate model compliance with human rights. | |
| GeoGR: Spatial-Semantic Token Compression and Group Routing for Efficient 3D VLMs Proposes spatial-semantic token compression and layer routing to accelerate 3D vision-language model inference across multi-view inputs. Ranked #351Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Presents token compression and dynamic layer-skipping techniques to accelerate VLM inference. | |
| MetaShare: Failure-Aware Collective Self-Improvement for Open-Ended Agents This paper introduces a framework for self-improving agents that records and shares both failure and success cases to prevent redundant mistakes. Ranked #352New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 17 Oct, 06:18 | On-topic Presents a multi-agent self-improvement architecture that shares failure and success strategies across tasks. | |
| OLiVES: An Outdoor Low-Light Video Benchmark for Enhancement and Segmentation This paper introduces OLiVES, a benchmark dataset of dynamic outdoor low-light videos designed for joint video enhancement and object segmentation. Ranked #352New testPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Introduces a multimodal video benchmark for joint enhancement and segmentation. | |
| SkillForgeBench: Evaluating Agent Reliability Grounded in Agent Skill Specifications This paper introduces a benchmark that uses agent skill specifications to evaluate the reliability and procedural correctness of AI agents. Ranked #352New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Presents a novel benchmark and rubric-based evaluation design for complex agentic tasks. | |
| inLoRA-MINT: Internal Representations for Auditing Training Data in Domain-adapted LLMs This paper presents an intermediate-representation method to audit whether specific private data was used in LoRA fine-tuning of language models. Ranked #352New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Audits private training data leakage in fine-tuned LLMs for AI governance compliance. | |
| Batch Before You Lift: Scalable Topological Deep Learning on Large Graphs Introduces Cluster-TNN, a mini-batch graph partitioning framework that enables scalable topological deep learning on large graphs. Ranked #352Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Introduces a scalable mini-batching method to overcome memory bottlenecks and speed up graph training. | |
| TEM: Temporal State Tracking Memory for Large Language Model Agents Presents a training-free framework that tracks changing conversational facts as versioned records to improve agent memory retrieval. Ranked #353New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 17 Oct, 06:17 | On-topic Introduces temporal state tracking memory specifically for multi-turn LLM agents. | |
| Look Before You Restore: Learning to Predict Before Acting for Agentic Image Restoration This paper proposes PreActIR, an agentic image restoration framework that forecasts tool consequences using a world model before taking action. Ranked #353New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Employs an agentic framework combining visual prediction and restoration tools. | |
| ATTCK-R: Knowledge-Aware Long-Horizon Jailbreaking for Protected LLMs This paper presents ATTCK-R, a planning-driven multi-turn red teaming framework to evaluate language models guarded by prompt safety defenses. Ranked #353New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Presents a red-teaming framework to evaluate and bypass prompt-based safety protections. | |
| WorkBorne: Discovering Agentic Tasks from Real-World Artifact Transitions This paper presents a framework for discovering and synthesizing terminal-based agent tasks grounded in real-world artifact transitions. Ranked #353New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Proposes a framework for synthesizing realistic benchmarks to evaluate agent execution. | |
| CAST: Credit Alignment for Cross-Tokenizer On-Policy Distillation Proposes a chunk-level credit assignment method for on-policy distillation across language models with different tokenizers. Ranked #353Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Addresses on-policy knowledge distillation across different tokenizers for efficient model training. | |
| LAMP: 3D Source-Aware Lighting Estimation at Multiple Points The paper proposes a framework using vision-language reasoning to estimate point-conditioned 3D directions and physical light source contributions. Ranked #354New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Fine-tunes vision-language models for 3D spatial lighting and physical source reasoning. | |
| Unchecked or Falsely Accused? Evaluating AI-Generated Text Detectors on Kazakh This work evaluates machine-text detectors on Kazakh text using pre-ChatGPT human writing and shows significant false-positive rates. Ranked #354New testPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Assesses the reliability and failure modes of deployed AI text detectors. | |
| Are Multimodal Benchmarks Really Useful? Item-Level Multimodal Benchmark Diagnosis via Structure-Response Co-Calibration The paper introduces a diagnosis framework to evaluate whether multimodal benchmark items genuinely require multi-image reasoning rather than unimodal shortcuts. Ranked #354New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Diagnoses multimodal benchmarks to detect shortcut items and evaluate benchmark validity. | |
| ChunkAct: Action Chunking for Agentic Robotics The paper introduces action chunking for agentic robotics to execute pre-planned primitive routines without redundant model reasoning at each step. Ranked #354New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Introduces action chunking to make LLM agents in robotics more efficient. | |
| DEX: Digit-Level Early Exit for Energy-Efficient MSDF Neural Network Inference Designs a most-significant-digit-first neural network accelerator supporting early exits and digit-level pruning for energy-efficient medical image segmentation. Ranked #354Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Proposes dynamic early exit and quantization to improve neural network inference efficiency. | |
| Multimodal AI Detection for Math-Heavy Scientific Text The paper examines AI-generated text detection in scientific papers by combining visual formatting, PDF text, and LaTeX representations. Ranked #355New testPreprint4/10 suggested | Multimodal models via arxiv · 17 Oct, 06:17 | On-topic Uses visual PDF layouts alongside text representations for multimodal document detection. | |
| Can LLM Agents Efficiently Identify Appliance Faults? Insights From Diagnosing a Simulated Appliance The paper evaluates the ability of language model agents to diagnose appliance faults using a simulated appliance benchmark based on sensor data. Ranked #355New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Introduces CoffeeBench to benchmark LLM diagnostic reasoning on simulated machinery. | |
| Generalizable Model Editing Could Be More Than Direct Generalization The authors analyze limitations in direct model editing benchmarks and compare them against observed generalization in practical chat settings. Ranked #355Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Examines reliable updating and factual behavior of deployed models. | |
| BLADE: Auditing Information Use for Failure Attribution in LLM Multi-Agent Systems The paper introduces an auditing framework that identifies failure steps in multi-agent systems by evaluating how information is utilized across handoffs. Ranked #355New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Develops a framework for diagnosing and attributing multi-step failure handoffs in LLM multi-agent systems. | |
| S$^3$R: Fast and Faithful RL Rollouts via Self-Sparse Speculative Decoding Introduces speculative decoding using self-sparse attention to speed up rollout generation during reinforcement learning. Ranked #355Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Addresses KV-cache memory traffic and efficient speculative decoding to accelerate model inference/rollouts. | |
| Perceive, Predict, and Plan: a Unified World Action Model for Egocentric Dexterous Manipulation Develops a unified world action model combining 3D perception, visual future prediction, and goal-directed planning for egocentric dexterous manipulation. Ranked #356New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 17 Oct, 06:18 | On-topic Develops a multimodal world action model integrating vision, language instructions, and 3D actions. | |
| Scaling Multilingual Safety and Reasoning Evaluation through Reasoning-Guided Translation The paper introduces a reasoning-guided translation framework to scale multilingual safety and reasoning benchmarks while preserving original task meanings. Ranked #356New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Develops a framework and collection for multilingual LLM safety and reasoning evaluation. | |
| 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 #356Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Evaluates execution reliability risks in automated coding-agent benchmarks. | |
| Autonomous Science Bench: Evaluating AI Agents for Autonomous Scientific Discovery The authors introduce a benchmark to evaluate autonomous scientific discovery agents making iterative experimental decisions under fixed budgets. Ranked #356New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates autonomous AI agents planning experiments under active experimentation budgets. | |
| SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Identifies distribution shift during autoregressive decoding and proposes a decoding-aware pruning method to accelerate model inference. Ranked #356Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Introduces decoding-aware pruning to accelerate LLM inference and reduce memory operations. | |
| Grounded Human-Attributed Description and Activity Recognition in Videos Introduces a benchmark and method to detect, describe, and recognize activities for every person visible in complex multi-person video scenes. Ranked #357New testPreprint4/10 suggested | Multimodal models via arxiv · 17 Oct, 06:18 | On-topic Improves vision-language models for grounded video description and activity recognition. | |
| Cross-Model In-Context Neurofeedback Reveals (Limited) Activation Control in LLMs The paper uses in-context neurofeedback sessions to evaluate whether language models can intentionally control their internal activations. Ranked #357Surprising resultPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Designs an evaluation method to test whether LLMs can intentionally control their internal activations. | |
| The AI's Poker Face: Contrasting Persuasion and Sycophancy in Voice-Based Multi-Turn Conversations The paper studies opinion change, rhetorical strategies, and sycophancy in spoken multi-turn dialogues between humans and language models. Ranked #357Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Investigates AI risks regarding user belief manipulation, sycophancy, and persuasion. | |
| AssetForge: Asset-Augmented Task Synthesis for Business Automation Agents AssetForge pairs reusable application assets with task rubrics to synthesize executable training tasks and evaluation checks for business automation agents. Ranked #357New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Presents a task-construction pipeline to train business automation agents with verifiable states. | |
| Time Delay System for Sequence Modeling Proposes a sequence modeling framework using dynamic nonlinear time delays based on semantic similarity between current and past inputs. Ranked #357Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Introduces a sequence architecture with O(1) per-step complexity for efficient processing. | |
| SurgStreamer: proactive, request-driven monitoring of surgical video streams with structured state and a reusable reasoning layer The paper introduces an event monitoring system for surgical video streams that decouples visual perception from procedure-agnostic logic reasoning. Ranked #358New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 17 Oct, 06:18 | On-topic Evaluates streaming vision-language models and video understanding in surgeries. | |
| MultiPartyTurnBench: Benchmarking Floor Control in Multi-Party Spoken Dialogue The paper introduces a benchmark to evaluate conversational floor control, turn completion, and interruptions in multi-party spoken dialogue systems. Ranked #358New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Introduces MultiPartyTurnBench, a new benchmark evaluating conversational floor control and turn-taking. | |
| Safety for Whom? Benchmarking Understanding Beyond Refusal in Egocentric Multimodal Assistants This paper introduces a benchmark to evaluate whether egocentric multimodal assistants understand who is harmed rather than solely tracking output refusal. Ranked #358New testPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Evaluates assistant safety failures and harm to users, bystanders, and society. | |
| Holding Agents Accountable with Verifiable Agent Insurance The authors propose verifiable agent insurance using smart-contract escrow to financially compensate users when agents violate programmatically logged behavioral promises. Ranked #358New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Addresses AI agent accountability and policy enforcement during multi-step tool execution. | |
| DPPM: Dual-Path Parametric Memory for Personalized Language Models The paper introduces a dual-path parametric memory method that combines pooled history representations with sequential adapter updates. Ranked #358Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Reduces inference context size by encoding long histories into lightweight LoRA adapters. | |
| MARS - A foundational map auto-regressor Introduces an autoregressive foundation model to generate vectorized road networks and building polygons directly from spatial inputs. Ranked #359New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 17 Oct, 06:18 | On-topic Introduces an auto-regressive model generating multimodal vectorized map primitives and visual geometry. | |
| 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. Ranked #359New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Proposes a meta-evaluation framework to evaluate the quality and efficiency of generated benchmarks. | |
| 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 #359New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Prevents reward hacking and ensures reliable verification of agent task execution. | |
| TextualRL: Textual Reinforcement Learning for Context Management TextualRL updates agent context prompts using groupwise trajectory comparisons rather than isolated feedback runs to improve policy adaptation without parameter tuning. Ranked #359New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Develops a reinforcement learning framework for managing context across agent rollouts. | |
| 4-Tensor Attention Model for Semantic Physical Reality The paper proposes a four-tensor attention architecture that jointly normalizes attention over semantic and temporal fibers for scene prediction. Ranked #359Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Demonstrates major speedups and lower compute requirements during model training and decoding. | |
| RIGen: Beyond Action Following with Reciprocal Interaction in Generative Gameplay Combines state-tracking rules with action-conditioned video generation to enable persistent reciprocal interactions in generative gameplay. Ranked #360New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 17 Oct, 06:18 | On-topic Focuses on multimodal action-conditioned audiovisual generation and interactive video simulation. | |
| Learning When to Trust Medical Knowledge: Regression-Safe Routing for Clinical Reasoning The paper introduces a benchmark for medical knowledge poisoning and a routing controller that selectively avoids unreliable retrieved evidence during clinical reasoning. Ranked #360New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Introduces a medical knowledge-poisoning benchmark to evaluate retrieval and reasoning robustness. | |
| Monarch-Muon: Structure-Aware Muon for Fast and Memory-Efficient Pretraining The authors propose an optimizer for Monarch-factored weights that replaces dense matrix orthogonalization with smaller block updates to reduce memory and pretraining time. Ranked #360Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Introduces a structured matrix optimizer that reduces memory footprint and training time for language models. | |
| 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 #360New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Improves reliability and safety by auditing and retracting contaminated agent memories. | |
| CompanionSafe: From Multi-turn Safety Evaluation to Guardrails for AI Companions This paper presents an evaluation benchmark and guardrail framework for detecting safety risks in multi-turn interactions with AI companions. Ranked #360New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates interactive conversational agents over long-term, multi-turn user relationships. | |
| Measuring and Reducing Unusable Replies of LLM Agents The paper measures whether requester users can understand multi-step agent replies and evaluates how explicit constraints improve reply usability. Ranked #361New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | On-topic Introduces a metric and reader-model benchmark to measure reply usability of LLM agents. | |
| When Sparsity Breaks Connectivity: Functional Reachability and Guided Topology Dynamics in Extreme Sparse Training The authors develop a reachability-guided dynamic sparse training method that preserves functional pathways during extreme network pruning and regrowth. Ranked #361Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Presents reachability-guided dynamic sparse training to enable extreme model pruning up to 99.9% sparsity. | |
| AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding The paper introduces an autonomous driving model unifying diffusion-based trajectory generation with autoregressive chain-of-thought reasoning to reduce latency. Ranked #361New way to see a problemPreprint3/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Introduces a unified vision-language-action model coupling visual understanding with action generation. | |
| Benchmarking and Training Reward Models for Cross-Source Evidence Grounding in Search Agents This paper introduces an evaluation framework and reward models to assess how well search agents ground generated answers in retrieved, independent sources. Ranked #361New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies how autonomous search agents ground evidence and resolve conflicting information during multi-step trajectories. | |
| Hide and Seek: HIDE Concepts to SEEK explanations for black box models Introduces a model-agnostic, concept-level attribution method for black-box vision models by masking concepts located with vision-language models. Ranked #361New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Provides faithful explanation and interpretability tools to audit black-box vision models. |
2,041–2,100 of 3,333← PreviousNext →