Every paper checked — nothing hidden
All papers
Every paper checked in the week of 12–18 Oct 2026 · LLM evaluation and benchmarks — nothing is hidden. Brief papers first.
All3,345In brief3More (on-topic)587Not relevant2,721Not assessed24Older45Could not confirm25Blogs with no paper54All blogs57
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| Divergence-Routed On-Policy Distillation for Language Models The paper proposes divergence-routed on-policy distillation to target critical token positions where students diverge from teacher language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Evaluates distillation on existing math benchmarks rather than proposing evaluation methods. | |
| Learning Across Temporal Graph Representation Silos The paper studies transferring learned predictive structure between continuous-time dynamic graphs and temporal knowledge graphs across distinct representations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Does not address language model evaluation or benchmarks. | |
| XferAgent: Cross-Agent Experience Transfer for Self-Improving LLM Agents The paper introduces an experience-graph framework enabling heterogeneous agents to share and reuse execution history across diverse tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on experience reuse mechanisms rather than benchmarking or evaluation methodology. | |
| Who's Behind the Text? Robust Model Authorship Attribution The paper proposes a likelihood-based attribution method to identify the source model of machine-generated text under distribution shifts. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents an attribution algorithm rather than an LLM evaluation benchmark. | |
| L3M: A Verifiable Harness and Runtime for Self-Extending and Self-Improving Agents The paper presents a Lean 4 runtime allowing self-extending agents to run verified computations across tools without constant human oversight. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on system runtime architecture rather than LLM evaluation benchmarks. | |
| JevCoder: When Does Decision Delegation Pay Off in Code Repair? The paper evaluates delegating code-repair decisions to low-cost probabilistic judges to reduce token costs without losing accuracy. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Applies existing SWE-bench benchmarks rather than developing new evaluation methodologies. | |
| Beyond Static Positional Encoding: Bayesian Time-aware Attention for Non-Stationary Long-term Time Series Forecasting The paper proposes a Bayesian formulation of self-attention to address non-stationary temporal dependencies in long-term time series forecasting. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents an architectural mechanism rather than LLM evaluation or benchmarks. | |
| Instruction-Conditioned Motion Control for Robotic Arms: Language-to-Code Decomposition and Reinforcement Learning The paper separates parameter specification, instruction generation, and code synthesis for language-conditioned robotic arm control using reinforcement learning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on robotic RL control rather than general LLM benchmarking methods. | |
| DAGFormer: Global Topology Prediction for Transformer Cross-Layer Connections The paper proposes a Transformer architecture that plans dynamic cross-layer connections globally across the model based on input tokens. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Evaluates an architectural proposal, not a new evaluation methodology. | |
| Eye, Robot: Learning Expressive Gaze Dynamics via a VLM-Guided Transformer-VAE The paper combines a vision-language model with a Transformer-VAE to generate expressive, human-like gaze dynamics for humanoid robots. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a robotics gaze controller, not an LLM benchmark. | |
| Sparse Autoencoders Are a Shared Coordinate System for Symbolic Matching The paper uses sparse autoencoders to establish a shared coordinate system for matching symbols across heterogeneous neuro-symbolic systems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Evaluates symbolic matching techniques, not general LLM benchmarks. | |
| Model What Matters: Decision-Guided Selective World Modeling for Language Agents The paper proposes a decision-guided co-training method that selectively filters observation units to improve world modeling for language model agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Proposes an agent training method, not an evaluation benchmark. | |
| AWARE-RG: Abstraction Aware Explanation of Resiliency Graphs for Cyber Physical Systems The paper presents a human-in-the-loop framework providing role-bounded explanations of cyber-physical system resilience graphs using Formal Concept Analysis. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a formal explanation system, not an LLM benchmark. | |
| HACHIMI-II: From Role-Play to Learner Behavior Modeling for Faithful Student Simulation The paper simulates student behavior by modeling individual learner history and constraining simulated actions via a third-person prediction harness. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a simulation technique for tutoring systems, not an LLM benchmark. | |
| MCAIR: Monitoring and Correcting Agent Unfaithfulness via Internal Representations The paper proposes a runtime monitoring method to detect and correct language model agents that deviate from intended procedures without reporting it. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a runtime monitoring probe rather than an evaluation benchmark. | |
| Function-Space Temporal Self-Distillation for Policy Optimization The paper introduces a single-policy optimization framework combining privileged adaptation with stored historical next-token distributions for language model reasoning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a reinforcement learning algorithm rather than an evaluation benchmark. | |
| PPAD-Completeness of Contiguous Envy-Free Cake Cutting with Additive Valuations The paper proves computational complexity results for contiguous envy-free cake-cutting and indivisible goods allocations under additive valuations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Theoretical computational complexity proof, not an LLM evaluation benchmark. | |
| C3: Rethinking Process Supervision via Confidence-Guided Credit Correction in Multi-Turn Reinforcement Learning The paper introduces a method that selectively adjusts positive credit for undesirable agent actions using the policy's confidence during multi-turn reinforcement learning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Proposes a training method, not a benchmark or evaluation framework. | |
| Runtime Monitoring of Supermartingale Certificates The paper proposes an algorithm using betting-based sequential tests to monitor supermartingale certificate violations during runtime execution of stochastic systems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Deals with formal stochastic runtime monitors, not LLM benchmarks. | |
| SceneAE: Scene Generation and Reconstruction in Queryable 3D-Anchored Latent Space The paper proposes a 3D-anchored latent space representation for simultaneous scene reconstruction and view generation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Introduces a generative latent architecture, not an evaluation benchmark. | |
| New Model, Old Habits: Behavior Transfer under Model Swapping in Multi-Session LLM Agents The paper analyzes model swapping in multi-session agents and proposes a history summarization method to preserve strong reasoning behaviors. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Examines multi-session agent behavior transfer rather than presenting evaluation methodology. | |
| Uncertainty-Driven Self-Play for Robust Multi-Agent Driving Policies The paper introduces an uncertainty-driven self-play method where exploring agents target policy weaknesses to learn robust autonomous driving policies. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Evaluates autonomous driving trajectories, not language model capabilities. | |
| Learning Individual Scientific Taste with Privileged Hindsight Distillation The paper models individual scientific taste using criterion assessments and introduces privileged hindsight distillation to align proposals with researcher preferences. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a distillation method for research taste, not benchmarking. | |
| ONCship: A Large-Scale Underwater Acoustic Corpus for Ship Recognition This paper introduces a large-scale underwater acoustic recording dataset paired with vessel tracking data for ship recognition tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Domain-specific acoustic corpus, not general LLM evaluation. | |
| Mitigating Video Hallucinations via Attention-Guided Representation Perturbation This paper presents a training-free contrastive decoding framework using attention-guided perturbations to reduce hallucinations in video language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Introduces a decoding framework, not evaluation methodology. | |
| What Happens to the Goal? Distinct Failures of Policy Alignment in Goal Misgeneralization This paper analyzes three distinct failure modes where reinforcement learning agents misgeneralize goals despite competent behavior. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Investigates policy alignment failure modes, not general LLM benchmarking. | |
| Exact Cut Complexity of Equal-Length Proportional Cake Cutting This paper establishes the exact cut complexity required for equal-length proportional cake cutting among multiple agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Theoretical cut complexity analysis, not an LLM benchmark. | |
| ProtModernBERT: Compute-Efficient Protein Language Modeling This paper introduces an efficient encoder-only protein language model incorporating modern architectural updates to lower training costs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Evaluates a specific model, does not propose new benchmarks. | |
| Does the Linguistic Register of Thinking Matter? Wenyan as a Compact Reasoning Register for Chinese LLMs This paper investigates using Classical Chinese as a compact intermediate reasoning register to reduce token length in Chinese language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Uses existing benchmarks, does not propose evaluation methods. | |
| PRISMATIC : Prescription Risk Inspection System for Multi-Agent Treatment Interaction in Clinical Decision-Making This paper proposes a multi-agent system designed to reduce clinical medication errors through dynamic collaborative coordination. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a multi-agent framework, not general LLM benchmarking. | |
| LiFi: LiDAR Generation from Multi-View Images via Geometric and Semantic Collaborative Guidance This paper presents a dual-branch latent diffusion method to generate LiDAR point clouds from multi-view camera images. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a generative 3D model, not an evaluation benchmark. | |
| One Agent per Zone: Decentralised Active Inference under Partial Observability for Multi-Zone HVAC Control on Edge Hardware This paper presents a decentralized active inference controller for multi-zone building heating and cooling systems on edge hardware. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents an HVAC control system, not an evaluation benchmark. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Introduces an architecture for video synthesis, not benchmarking methodology. | |
| Learning Perception to Action Rules through Reinforcement Learning for Theory-of-Mind Reasoning This paper trains language models using reinforcement learning to generate explicit perception-to-action rules for theory-of-mind reasoning tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Uses benchmarks for experiments, does not propose evaluation methods. | |
| VaryST: Generalizable Spatio-Temporal Forecasting Beyond the Closed-World Setting This paper presents a compact spatio-temporal forecasting model designed to generalize across shifting spatial domains and varying observation availability. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a forecasting model architecture, not evaluation methodology. | |
| One Monitor Does Not Fit All: Formalism-Agnostic Runtime Verification for Agentic AI This paper develops a runtime verification architecture that decouples monitoring formalisms from execution boundaries to safely control AI agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a runtime monitoring architecture, not an LLM benchmark. | |
| ActionOPD: Distilling Decisions, Not Tokens This paper proposes an on-policy distillation method that aligns student agents using semantic actions rather than token-level supervision. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Introduces a training distillation method, not an evaluation framework. | |
| Teaching Agents in their Own Style: Blackbox Teacher Distillation for Web Agents This paper presents a distillation method where a teacher model minimally edits a student web agent's reasoning traces at failure points. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focus is distillation methodology, not evaluation benchmark design. | |
| LAGO: Language Agents with Reusable Growing Options This paper introduces an architecture that stores and reuses specialized language agents as temporally extended options for complex coding tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on an agent framework evaluated on existing benchmarks, not new evaluation methods. | |
| Memory-Efficient Subspace Optimization with Residual Correction: From Momentum SGD to Muon This paper presents a memory-efficient subspace optimization framework using residual correction to stabilize language model training. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Develops training optimizers, not model evaluation frameworks. | |
| DeltaStreamer: Towards Human-Like Agentic Streaming Video Understanding This paper proposes an agentic video framework combining change-aware memory and targeted replay for streaming video understanding. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Proposes a streaming video method, not an evaluation benchmark. | |
| Euler: Towards Human-Centered Multi-Agent Mathematical Discovery This paper introduces a multi-agent mathematical discovery system that organizes proofs, assumptions, and failed attempts into structured research graphs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents a multi-agent mathematical framework rather than a general evaluation benchmark. | |
| Explorer-Zero: Improving Self-Play for Long-Horizon Tasks with Decomposable Verifiable Rewards This paper proposes a self-play reinforcement learning framework that automatically generates tasks and intermediate verifiable rewards for long-horizon agent training. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Proposes an RL training technique, not an LLM evaluation benchmark. | |
| Rationale Distribution Distillation via Flow Matching for Distributional Continuous Latent Reasoning This paper presents a flow matching method to distill diverse continuous latent reasoning distributions from multiple chain-of-thought rationales. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Introduces a distillation method rather than an evaluation benchmark. | |
| Neural Wasserstein Splines: Representation-Invariant Second-Order Priors for Trajectory Inference The paper presents a continuous trajectory inference method that regularizes population dynamics using neural splines and Wasserstein smoothness. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Develops a mathematical inference algorithm, not an evaluation method. | |
| Frustratingly Easy Majority Voting Using Length Conditioning The paper proposes weighting majority voting responses using a length-conditioned prior adapted at test time without model fine-tuning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Presents an inference aggregation algorithm rather than an evaluation benchmark. | |
| DeepPV-MRF: End-to-End Partial-Volume Fraction Estimation from MR Fingerprinting Time Series The paper introduces a deep learning framework to estimate voxel-wise brain tissue fractions directly from magnetic resonance fingerprinting signals. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Evaluates an MRI reconstruction model, not language model benchmarking. | |
| Universal Sample-Based Embeddings of Higher-Order Objects The paper proposes a sample-based neural framework to learn fixed-size embedding representations for higher-order objects like distributions and graphs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on representation learning, not LLM evaluation or benchmarking. | |
| Multimodal Medical Image Parsing with Task-Specific Low-Rank Adapters The paper describes a vision-language model using task-specific low-rank adapters for multimodal medical image parsing. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Describes a competition submission architecture rather than general LLM evaluation methodology. | |
| MUCO: A Mani-Unit-Structured Manipulation Framework for Compositional Long-Horizon Robot Tasks The paper proposes a modular manipulation framework that sequences verified robot capability units for long-horizon tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Proposes a robot manipulation framework rather than an evaluation benchmark. | |
| Does Machine Translation Retain Media Bias? The paper investigates whether machine translation systems preserve or neutralize media bias across languages. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on translation bias retention, not LLM evaluation benchmarks or methodologies. | |
| Learning Graph Energies for Finite-Time Generation The paper proposes an energy-based learning method that optimizes acceptance rates for finite-step discrete graph sampling. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Proposes a generation technique rather than an evaluation benchmark or judge framework. | |
| Quiet Triggers, Loud Footprints: A Tri-Space Measurement Study of Backdoor Stealthiness The paper presents an empirical framework measuring neural network backdoor stealthiness across input, feature, and parameter spaces. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Analyzes backdoor stealthiness metrics in vision, not general LLM evaluation methods. | |
| AtomSearch: Evidence-Grounded Tree Search for Agent Skill Optimization The paper proposes a tree search method that optimizes agent skills through contrastive atomic rule editing and execution memory. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Main contribution is an optimization framework, not a benchmark or evaluation method. | |
| Neurosymbolic Learning of Interpretable World Models The paper introduces a neurosymbolic world model that learns interpretable concepts from temporal logic constraints for agent planning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on learning world representations rather than LLM benchmarks. | |
| Safe-FCL: Continual Risk Awareness and Fine-Grained Safety Aggregation for Federated Continual Learning of LLMs The paper proposes a federated learning framework with risk-aware aggregation to prevent malicious updates from compromising global model safety. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Proposes a defense framework rather than LLM evaluation benchmarks. | |
| Empirical Boundaries of Reservoir Computing for Few-Shot Sequence Pattern Tasks The paper empirically analyzes echo state network variants on synthetic sequence-to-sequence pattern tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Studies recurrent neural net limits rather than LLM benchmarks. | |
| What Coding Agents Leave Unsaid: Measuring Task-State Gaps Between Execution and Reports The paper measures the information gap between coding agents' internal execution logs and their user-facing progress reports. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on agent oversight gaps rather than general benchmarking. | |
| Provable Transformer Context Compression for In-Context Reinforcement Learning The paper analyzes the minimum context size required for transformer models to perform in-context reinforcement learning without parameter updates. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Focuses on theoretical bounds rather than benchmark evaluation. | |
| On the Convergence of Semi-Gradient TD: A Mean Field Perspective The paper provides a theoretical analysis of stability and convergence for semi-gradient temporal-difference learning using mean-field neural networks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06 | Not relevant Mathematical proof paper rather than LLM benchmarking. |
1,381–1,440 of 3,345← PreviousNext →