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 |
|---|---|---|---|
| PACT: A Pragmatic Prompting Framework for Expressive and Voice-Preserving Speech-to-Speech Translation Presents a pragmatic prompting framework to preserve speaker voice, emotion, and tone during speech-to-speech translation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Does not propose evaluation benchmarks or benchmark analysis. | |
| PAINT: A Framework for Multilingual Pain Representation from Medical Licensing Exams to Real-World Clinical Texts Presents a multilingual framework and dataset spanning four languages to evaluate clinical pain representations across exams and clinical notes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a medical domain dataset rather than general LLM evaluation. | |
| Prompt or Policy? Optimizing Physician Slate Presentations with LLM User Simulators Develops a simulation framework using persona-guided models to optimize comparative doctor summary presentations for medical patients. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Uses LLM judges as a component, but does not benchmark LLMs. | |
| Jev-Guard: Fast Routing and Auditing Agent Actions over Editable Policy Trees Introduces a low-latency guardrail model that routes incoming agent actions through editable policy trees for safety auditing. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on deploying guardrail routing rather than benchmark design. | |
| Probing the Geometry of Color Representations in Multilingual Language Models Probes layer representations in multilingual language models to study whether internal color geometry aligns with human perception or language syntax. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on representation probing, not benchmarking or evaluation methodology. | |
| Pin It on Anyone: Certified Targeted Displacement of Multi-Bit LLM Watermarks Without the Key Demonstrates an attack method that alters embedded multi-bit text watermarks without knowing the secret generation key. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates watermarks rather than general LLM capabilities or benchmarks. | |
| Multi-Stream Residual Computation in Diffusion Language Models Equips diffusion language models with multiple residual streams and adaptive routing to separate visible context from masked token generation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Does not propose evaluation frameworks or benchmark analyses. | |
| Prompt-Injection Probes Learn the Conversation, Not Just the Injection Shows linear prompt-injection monitors overfit to conversation features and reduces their false alarms by adding matched harmless negative examples. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on detection defense mechanisms rather than benchmark development. | |
| Revisiting the Estimation of Unigram Surprisal in Computational Psycholinguistics Proposes estimating both contextual and unigram surprisal from the same language model to reduce estimation mismatch in reading time studies. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on statistical psycholinguistic estimators rather than LLM benchmarking. | |
| Reliable Memory from Unstable Growth: Curating Self-Written Agent Playbooks with Randomized Experiments Presents a randomized experiment framework to test and curate self-written agent playbooks, preventing task failure caused by unverified lessons. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an agent memory curation method rather than an evaluation benchmark. | |
| Estimating Surprisal from the Speech Signal Estimates surprisal directly from continuous speech signals using a speech language model and importance sampling over abstract linguistic units. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an estimator for psycholinguistic surprisal, not LLM benchmark evaluation. | |
| Copying Is Not Grounding: Training Faithful Long-Form RAG with Authored-Sentence Credit Identifies passage copying as a failure mode in long-form retrieval systems and introduces sentence-level credit assignment to encourage genuine grounding. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a training method for RAG, not a general benchmark. | |
| Teacher-Guided Tool Learning for Small Language Models via Bidirectional KL Constraints Proposes a teacher-guided reinforcement learning method using bidirectional constraints to help small language models learn effective tool use. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a training algorithm, not an evaluation benchmark or judge framework. | |
| Imprinting: Knowledge Distillation Can Shift Student Behavior Demonstrates that knowledge distillation transfers safety dispositions and behavioral tendencies from teacher models to student models on unrelated tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes distillation side-effects rather than benchmarking or evaluation metrics. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a steering method for retrieval, not an evaluation framework. | |
| Legal-R1: Agentic Retrieval for Evidence-Based Legal Reasoning This paper introduces a multi-agent retrieval framework that coordinates search across legal statutes and precedent cases for grounded legal reasoning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an agentic RAG system rather than a new general evaluation framework. | |
| Breakdown of Conversational Involvement: A Joint Multi-Modal Analysis of Conversation Floors and F-formation Detection with Multiple Features This paper analyzes multi-person conversational involvement by jointly evaluating speaking floors and spatial formations from multimodal video features. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies social conversation floor detection rather than language model evaluation. | |
| Bayesian Test-Time Inference of Task-Aligned Similarity from Weak Interactive Feedback This paper proposes a Bayesian method that infers task-aligned similarity metrics at test time using weak interactive feedback. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a Bayesian test-time inference method, not evaluation benchmarks or methodology. | |
| Code-Switching Detection via Language-Selective Dimensions in Speech Foundation Models This paper presents a method to detect spoken code-switching directly from speech representations using language-selective hidden dimensions. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a speech detection method rather than LLM benchmarking methods. | |
| On the Monitorability of Natural Language Autoencoders: Identification vs Discrimination This paper uses referential communication games to analyze why natural language autoencoders fail to provide discriminative oversight of model activations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes activation autoencoder theoretical oversight, not LLM benchmark evaluation. | |
| SkillScaffold: Scaffolding Agentic Reinforcement Learning with Policy-Adaptive Skill Credit Assignment and Selective Injection This paper introduces a reinforcement learning scaffold that selectively injects textual skills into agents using low-cost utility estimation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a training scaffold algorithm rather than an evaluation benchmark. | |
| The Last Remote Sensing Model Built by Humans? Casting Any Earth Observation Task as Recursive Self-Improvement This paper frames Earth observation model development as an automated loop where coding agents build and iteratively refine models using verifiable harnesses. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on training workflow automation rather than benchmark methodology. | |
| Embodied Zero-Shot Task Planning via Hardware-Aware LLM-VLM Integration for Grounded Robotic Manipulation This paper presents a robotic manipulation framework combining high-level language planning with vision-language verification to enforce physical and kinematic constraints. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an embodied manipulation framework rather than an evaluation benchmark. | |
| ClawLoop: Self-Improving Agents via Harness-Native Training Loop This paper introduces a training framework that enables harness-based software agents to self-improve without relying on external teacher models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a training framework rather than evaluating benchmark saturation or LLM-as-judge. | |
| Correction Training for Diffusion Models This paper introduces a learned correction step to reduce error accumulation and the training-inference gap in iterative diffusion model sampling. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a training method for diffusion models, not an evaluation benchmark. | |
| Beyond Discrimination: Evaluating Probability Calibration in Radiomics-Image Fusion for Renal Cell Carcinoma Classification This paper evaluates whether combining radiomics with deep image representations improves probability calibration and discrimination in renal cancer classification models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates probability calibration on medical datasets, not LLM evaluation benchmarks. | |
| Beyond Gate Weights: Decision-Level Interpretation of Radiomics-Image Feature Fusion for RCC Classification This paper analyzes decision logits in radiomics-image fusion models to distinguish gate weight values from actual feature contributions during cancer classification. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies feature interpretability in medical fusion models, not LLM benchmarks. | |
| Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation This paper introduces a coarse-to-fine text generation method using discrete code hierarchies to accelerate parallel sampling over diffusion language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a generative modeling architecture rather than benchmark evaluation methods. | |
| Visualizing the World Model in an Agent's Value Function The paper uses counterfactual interventions to visualize and examine implicit environmental concepts learned inside a reinforcement learning agent's value function. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies world representations in RL value functions, not LLM benchmarks. | |
| Cross-Modal Latent Influence: Patch-Level Training Sample Attribution for Vision-Language Models The authors introduce sparse autoencoders to trace multimodal training influence down to specific image regions and text phrases. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on internal latent attribution methods rather than benchmark evaluation. | |
| Should SLMs be in the Driver's Seat? A Study of Local Models as Interpreters in Cloud-Written Agent Programs The paper demonstrates a hybrid architecture where cloud-generated agent code delegates localized private-data queries to edge models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates agent architectural design rather than general LLM benchmarking. | |
| Cortex: Measurement-Gated Evolution and Pruning of Coding-Agent Skills and Rules The authors develop a system that systematically tests, evolves, and prunes skill and rule files used by coding agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Tool for optimizing agent harnesses rather than general LLM benchmarking. | |
| Learnable Low-Rank Polynomial Sketch for Effective Linear Attention The paper proposes a learnable low-rank polynomial sketch to reduce approximation errors in linear attention mechanisms. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes attention approximations rather than LLM evaluation benchmarks. | |
| Harness^2: Recursive Agent Harnessing for an Open World The authors propose a two-level recursive self-improvement framework to evolve and refine execution harnesses for open-world agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on recursive agent improvement rather than evaluation benchmarks. | |
| CAPTOR: a vision-language foundation model for Chest-Abdomen-Pelvis Oncologic CT The paper develops and evaluates a multimodal foundation model for interpreting oncologic chest, abdomen, and pelvis CT scans. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates medical foundation models rather than LLM benchmarks. | |
| Latent World Action Model for GUI Agents The paper presents a latent world action model that predicts action-induced state transitions rather than full-screen pixel reconstructions for graphical agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an agent architecture rather than LLM evaluation benchmarks. | |
| Agentic Data Engineering for LLMs: System Design and a Controlled Empirical Study The authors present a framework where autonomous agents iteratively refine data engineering strategies to improve model training outcomes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies agentic data generation strategies rather than LLM evaluation benchmarking. | |
| Akule: Fast and Scalable Multi-Robot Motion Planning via Sparse Interaction Diffusion The paper introduces a sparse interaction diffusion method that coordinates multi-robot trajectories using a frozen single-robot prior and modular corrections. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates robotic trajectory planning rather than LLM evaluation benchmarks. | |
| Echo: Error-Compensated 4-Bit Tensor-Parallel Communication for LLM Training and Inference The authors develop an error-compensated 4-bit all-reduce algorithm to reduce communication overhead during distributed model training and inference. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on tensor-parallel communication speed rather than LLM benchmarks. | |
| The RNN Homunculus: Learning to Gate Long-Term Memory The paper studies gate learning in recurrent networks, showing how adding short-dependency samples helps models learn long-range temporal connections. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies gated RNN sequence learning rather than LLM benchmarking. | |
| GeoFix: Geometry Fixing for 4D Consistent Generation The authors introduce a framework that uses cross-view depth consistency as a single reward to improve 4D video generation geometry. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on generative consistency rather than LLM evaluation benchmarks. | |
| Controllable Diverse Sampling through Pullback Directions The paper proposes a training-free method leveraging pullback directions to increase sample diversity in conditional diffusion generation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Investigates diffusion sampling diversity rather than LLM benchmarks. | |
| RUBATO: Decoupled Clocks Allow Full-Duplex Models to Think as Fast as They Can The authors propose decoupled positional encodings enabling full-duplex speech models to run internal thinking streams at hardware-adaptive speeds. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Does not benchmark language models or evaluate evaluation methodology. | |
| ConicFlow: constrained Inference-time optimal control for flow matching models The paper introduces a receding-horizon optimal control algorithm for constrained sampling in flow matching generative models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents sampling control methods, not LLM evaluation benchmarks. | |
| MvFFN: Multi-view Floor plan Feed-forward Network for Unposed Wide-Baseline Panorama Layout Reconstruction The authors present a feed-forward network that aggregates local wall predictions to reconstruct floor plans from sparse unposed panoramas. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates 3D layout prediction rather than LLM benchmarks. | |
| EV-JEPA: Elastic V-JEPA for Any Device and Cross-Device Inference The paper presents an elastic vision architecture that dynamically reconfigures width and supports partitionable inference across heterogeneous edge devices. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Does not propose benchmarks or evaluations for LLMs. | |
| TraceSampler: Making Sampling Errors Predictable and Correctable for Accelerated Generative Sampling The paper introduces a post-training residual correction method to predict and correct endpoint errors during low-step generative model sampling. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes sampling error correction, not LLM benchmark evaluation. | |
| Which Metadata Matters for LLM Tool Selection in MCP? Evidence from Controlled Factor-Level Interventions The authors run controlled experiments showing how parameter schema differences in metadata alter language model tool selection. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates tool metadata impact rather than general LLM benchmarking. | |
| Natural Code Transformation Confuses Agents in Vulnerability Analysis The study reports that natural code transformations confuse vulnerability analysis agents, proposing functional changes as a potential defensive strategy. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Explores agent attack defense rather than general LLM evaluation benchmarking. | |
| Actionable Meta-Requirements for AI Robustness Auditing The paper establishes and validates a set of meta-requirements to guide the design of frameworks for auditing AI system robustness. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Addresses governance auditing requirements rather than LLM benchmark evaluation methodology. | |
| DexFLEX: Contact-Aware Foundation Controller for Command-Guided Dexterity The authors present a foundation controller that converts motion drafts into contact-consistent joint commands for dexterous robotic hands. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a robotic motion controller rather than an evaluation benchmark. | |
| ANASSA: From High-Level Framework to Low-Level Architecture for Spatial Agentic AI The paper refines a high-level spatial AI orchestration framework into an implementation-focused low-level architecture for governed agent execution. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a system architecture design rather than an evaluation benchmark. | |
| Deep learning from a control vantage point The paper reviews deep learning architectures and training dynamics through the lens of continuous-time nonlinear control theory. Peer-reviewed | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Theoretical overview of architectures, not benchmark evaluation or testing methodology. | |
| SAT-RL: Training Language Models for Logical Reasoning with Symbolic Solvers The paper presents a reinforcement learning framework that trains language models to solve complex logical reasoning tasks using interactive symbolic solvers. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a training method with solvers rather than an evaluation framework. | |
| Memory Without Curation: Learning Multi-Stream Latent Memory for Robot Policies The paper proposes an uncurated multi-stream latent memory framework that augments vision-language-action robotic policies while maintaining low inference latency. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on a policy architecture rather than general LLM evaluation benchmarks. | |
| Coarse-Graining Recursive Expert Computation: Continuation-Relative Interfaces and Failure Modes The paper introduces an evaluation framework to measure the trade-off between task quality and computation cost when replacing iterative neural modules. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Does not introduce a standard benchmark or LLM evaluation dataset. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on a deployment case study rather than a general evaluation benchmark. | |
| Target-ESS Tempered Policy Optimization (TETPO): Smooth Variance Control via Bridge Importance Sampling The paper proposes a data-adaptive importance sampling method to stabilize variance and prevent sample size collapse in asynchronous RL post-training. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an RL optimization algorithm rather than an evaluation benchmark. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Does not introduce general LLM evaluation benchmarks or methodologies. | |
| Jevolution: A Framework for Studying Communication and Collective Dynamics in AI Agents The paper presents an instrumented predator-prey simulation framework to evaluate communication, scheduling, and collective dynamics in remote decision-making agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a simulation framework for agent communication rather than general LLM evaluation. |
2,101–2,160 of 3,345← PreviousNext →