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 |
|---|---|---|---|
| On Symmetry for Characterizing Agent Heterogeneity in Multi-Agent Systems The paper defines an algebraic symmetry framework for characterizing structural agent heterogeneity and policy representations in multi-agent systems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Mathematical analysis of agent symmetry, not an LLM benchmark. | |
| Input-Conditioned Search for Efficient Test-Time Scaling of Recursive Reasoners The paper presents a test-time search method that perturbs input conditions to generate diverse reasoning paths in recursive reasoning models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an inference scaling technique, not an evaluation benchmark. | |
| Configuring RAG Systems: What Breaks, Where Originates, and How to Fix The paper presents an empirical study of 654 real-world configuration bugs across ten retrieval-augmented generation systems, categorizing causes and fixes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies software configuration bugs in RAG rather than LLM benchmarking. | |
| Order-Consistent Latent Geometry for Reliable Short-Horizon Planning in World Models The paper analyzes pathwise distance consistency in visual world models to avoid candidate order reversals during latent short-horizon planning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Theoretical planning geometry paper, not an LLM benchmark. | |
| LLM Agents for Automated Post-Training via Structured Recipe Optimization The paper introduces an agent-based approach that adapts search spaces to optimize post-training recipes and configurations for pre-trained language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on automated training recipe search rather than benchmark evaluation. | |
| The Many Morals of AI: The cautionary bias of North American news narratives in global perspective The paper analyzes implicit morals in news coverage of artificial intelligence across four global regions, identifying distinct regional variations in narrative caution. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes news discourse rather than evaluating LLM capabilities or benchmarks. | |
| When More Objectives Make Identification Easier: The Role of Preferences and Feedback The paper theoretically analyzes multi-objective best-arm identification to determine when multiple criteria reduce exploration requirements in multi-armed bandit settings. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies theoretical multi-objective bandits, not language model evaluation. | |
| Compare to Correct: Pairwise-Guided Logit Adjustment for LLM Test-Time Scaling The paper introduces a logit adjustment method based on pairwise comparisons between sampled reasoning paths to improve test-time answer selection in language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a decoding adjustment method rather than an evaluation framework. | |
| GeoMER: A Joint Euclidean-Hyperbolic Geometry Framework for Multimodal Emotion Recognition The paper presents a multimodal emotion recognition framework that integrates Euclidean and hyperbolic geometry to capture hierarchical emotion structures from text and speech. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a representation learning method, not an LLM evaluation. | |
| ACCRUE: Continual Off-Policy Reinforcement Learning for Robot Foundation Models The paper proposes a continual off-policy reinforcement learning framework that reuses actor and critic networks across sequential robotic manipulation tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a robotic RL algorithm, not an LLM benchmark. | |
| PhysicsSliders: Learning Physical Controls for Audio Foley Generation The paper presents a generative audio method that controls Foley sound synthesis using physical interaction, material, and geometry parameters. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates audio generation physics, not general LLM benchmarking. | |
| Rethinking Projector Training in Multimodal LLMs The paper examines projector training in multimodal models and proposes a training-free initialization method using summary statistics instead of paired image-caption data. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an alignment method rather than evaluation benchmarks or methodology. | |
| Auditing Future-Memory Use in Point-in-Time Language Agents The paper audits historical evaluation tasks to see whether an agent's internal post-cutoff memory skews its point-in-time evidence gathering. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Audits a specific model behavior rather than introducing evaluation benchmarks. | |
| LLM2R: From Large Language Models to 3D Reassemblers The paper adapts a causal language model using compressed 3D geometric tokens to assemble multi-part objects in a single forward pass. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a model architecture for 3D assembly, not evaluation benchmarks. | |
| Beyond Answer Voting: Fine-Grained Correction of LLM Reasoning through Graph Alignment across Sampled Traces The paper proposes a framework that uses reasoning graphs to locate claim-level disagreements across sampled traces and correct multi-step language model reasoning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an inference-time verification method rather than benchmark evaluation. | |
| DPOison-VL: Clean Preference Poisoning in Vision-Language Direct Preference Optimization The paper introduces an imperceptible image perturbation attack that manipulates alignment in vision-language models during preference optimization without altering text annotations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Develops an attack method rather than evaluation frameworks or general benchmarks. | |
| Nudge the Judge: Uncovering Program Decision Hijacking in Real-World Jev Applications The paper presents an automated framework to evaluate how adversarial contexts can redirect program execution through valid structured judgments in language-model applications. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a vulnerability testing framework for specific apps rather than LLM benchmarks. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates an agent system on benchmarks rather than introducing evaluation frameworks. | |
| Two Routes to Interoceptive Prioritisation: Dissociating Likelihood Precision and Preference Reweighting in Active Inference The authors compare information precision and preference reweighting mechanisms in active-inference agents managing competing survival needs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies active inference simulations rather than LLM benchmarks or evaluation. | |
| Falsification-proof non-market allocation mechanisms The paper characterizes optimal falsification-proof non-market allocation mechanisms when individuals can strategically manipulate their observable characteristics. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Theoretical economic study with no connection to LLM benchmarks or evaluation. | |
| Geometric Calibration of Supervised Fine-Tuning This paper presents a post-hoc method that refines fine-tuned models by selectively reverting parameter updates along principal directions of base weights. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a fine-tuning method rather than evaluation benchmarks or methodology. | |
| DTScribe: Generating Models for Digital Twin Reporting using Local Large Language Models The authors present an approach using local language models to automatically generate conceptual architectural models and reports for digital twins. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Uses LLM-as-a-judge for an application rather than studying evaluation methodology. | |
| ST-GRU PredRNN-V2: A GRU-Based Reformulation of PredRNN-V2 for Video Prediction The authors reformulate a recurrent video prediction architecture using gated recurrent units to reduce parameters while maintaining spatiotemporal modeling. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Applies a model to existing benchmarks rather than building evaluation methods. | |
| Tiny Recursive Message Passing This paper applies recursive training and discrete deduction procedures to message-passing neural networks to solve constraint-reasoning tasks with few parameters. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates an architecture on puzzles rather than proposing evaluation benchmarks. | |
| Kwai Summary Attention V2 Technical Report This technical report presents an updated summary-token attention architecture to optimize cache compression and computation across model layers. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a model architecture rather than evaluation benchmarks or methodology. | |
| Known-Voice Cloning and the Limits of Personality Transfer for Character Design This paper investigates whether cloning a familiar voice systematically transfers personality impressions to an embodied virtual character. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant User study on voice cloning perception, not an LLM benchmark. | |
| SMI: Ontology-Grounded Semantic IDs for Hierarchy-Aware Medical Concept Representation The authors propose a training-free method constructing ontology-grounded semantic identifiers to preserve hierarchy in medical concept embeddings. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an embedding representation framework, not an evaluation benchmark. | |
| From Relation-Specific Heads to Shared Interventions: Contextual Entrainment in Language Models This paper investigates attention heads responsible for contextual entrainment and evaluates whether their ablation transfers across different semantic relations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on interpretability and attention head ablation, not benchmark design. | |
| Model-Based Online Decision Making via Generative Trajectory Planning The authors introduce a decision-making framework that samples candidate action sequences from a diffusion model and refines them using learned dynamics. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a control planning algorithm, not an evaluation benchmark. | |
| ProAS: A Progress-Aware Advantage Shaping Framework for Multi-Turn Agentic Reinforcement Learning This paper proposes a progress-aware credit assignment framework for multi-turn reinforcement learning in language model agents without using critic models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an RL training algorithm for agents rather than an evaluation benchmark. | |
| RadarGSOcc: Doppler-Aware 3D Semantic Occupancy Prediction Using 4D Radar Tensor with Geometric-Semantic Feature Extraction This paper proposes a framework and benchmark for 3D semantic occupancy prediction in autonomous driving using four-dimensional radar tensors. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a perception benchmark for autonomous driving, not LLMs. | |
| Just Respect the Semantics: Image-Token Compression via Pre-LLM Semantic Variable-Resolution Tokenization and In-LLM Relevance Pruning The authors present an image-token compression technique using variable-resolution merging for semantic regions and learned relevance pruning inside the model. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Compression technique paper rather than an evaluation benchmark. | |
| GeoReader: Learning Latent Geometry for Spatial Reasoning in Vision-Language Models This paper introduces a framework that trains lightweight readers to convert latent 3D geometric representations into tokens readable by vision-language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a model architecture for spatial QA, not an evaluation benchmark. | |
| UniW(A)M: A Unified Mixture-of-Experts Framework for World and World-Action Models The authors propose a mixture-of-experts diffusion framework that unifies world prediction and action generation models across shared physical dynamics. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Architecture for robot action models rather than LLM benchmarks or evaluation. | |
| Unlocking Token Rewards via Training-Free Reward Attribution This paper presents a training-free, gradient-based method to extract token-level reward signals directly from process reward models. Peer-reviewed | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on RL reward modeling algorithms rather than evaluation benchmarks. | |
| Steertuning Policies from a Few Human Videos The authors introduce a system that enables robots to learn new tasks from five human video demonstrations via fast policy finetuning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a robot policy tuning method rather than an LLM evaluation benchmark. | |
| A Rate Transfer Method: Convergence Rates in Stochastic Gradient Descent via Stochastic Calculus This paper presents a theoretical rate-transfer method using stochastic calculus to establish finite-time convergence bounds for stochastic gradient algorithms. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Mathematical optimization analysis rather than LLM benchmarking or evaluation methods. | |
| PreciseCoder: Faithful and Generalizable Debugging Beyond Unit-Test Rewards The paper introduces a debugging training algorithm that rewards localized semantic code edits alongside unit test execution. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a post-training technique for code models rather than an evaluation benchmark. | |
| Jeva: System One Models from First Principles The paper proposes an alternative architecture for fast decision-making that handles missing choices, option permutation, and candidate scaling natively. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a decision architecture rather than benchmarking or evaluating existing LLMs. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a diagnostic attribution framework, not a general evaluation benchmark. | |
| Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents The paper presents a runtime system that structures long-horizon agent memory around reusable progress objects rather than raw interaction history. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an agent context runtime rather than an evaluation benchmark. | |
| Bytes to Threats: Byte-Sequence Transformer For Malware Detection The paper presents a byte-level transformer model designed to detect malware directly from raw executable byte sequences. Peer-reviewed | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a malware classifier rather than an LLM evaluation framework. | |
| Optimizing Team Behavior via Extremum-seeking Control in Multi-agent Reinforcement Learning The paper introduces an adaptive diversity control framework that optimizes cooperative multi-agent reinforcement learning teams using episodic return feedback. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a control framework for RL, not an LLM benchmark. | |
| Self-Evolving Agent Harness The paper formalizes a framework and theoretical generalization bounds for evolving an agent's software harness without updating model weights. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes harness optimization theory rather than proposing an LLM evaluation benchmark. | |
| Back-Translation Still Matters: Leveraging Monolingual Data for Low-Resource Machine Translation with LLMs The paper examines back-translation using monolingual data as an alternative to continued pretraining for low-resource translation with language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on low-resource translation training, not benchmark design or LLM evaluation. | |
| Read the Room Before Your Model Generates The paper introduces a decoding framework that dynamically selects sampling temperature based on geometric representations of conversational scenarios. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on a decoding generation method rather than evaluation frameworks. | |
| RoboSimStudio: An Interactive Toolkit for Rapid Real2Sim of Robot Manipulation Tasks The paper presents an interactive simulation toolkit that enables humans and code agents to iteratively calibrate robot manipulation physics. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents simulation software, not LLM evaluation methods or benchmarks. | |
| Fast Rates for Offline Q-Learning with KL Regularization The paper analyzes sample complexity rates for offline Q-learning algorithms using Kullback-Leibler regularization. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Mathematical proof of Q-learning bounds, not LLM evaluation or benchmarks. | |
| Who Improves the Robot? Toward Physical Recursive Self-Improvement The paper analyzes physical recursive self-improvement in robotics, proposing a multi-level evaluation framework and reviewing existing adaptation methods. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Conceptual framework survey, not an empirical benchmark for language models. | |
| MFG: Multi-turn Feedback Guided Molecular Optimization The paper presents a reinforcement learning framework that trains language models to optimize molecules across multiple turns using iterative feedback. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Applies RL to chemistry tasks rather than general LLM evaluation benchmarks. | |
| Two Clocks: Why Larger Rollout Batches Learn Slower but Fail on Schedule in Reinforcement Learning with Verifiable Rewards The paper develops a framework explaining why large rollout batches in reinforcement learning reduce sample efficiency despite maintaining drift thresholds. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Investigates optimization theory rather than benchmark design or LLM evaluation. | |
| Hybrid Reasoning Agents The paper proposes an agent architecture that dynamically switches between latent reasoning and text interaction while maintaining external tool access. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates on existing benchmarks rather than proposing an LLM evaluation framework. | |
| AesCode: Aesthetic Code Generation with Decoupled Cross-Modal Rewards The paper presents a method for generating visual artifact code by combining text prompts with generated image guidance and cross-modal rewards. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on cross-modal code generation rather than standard LLM evaluation methodologies. | |
| DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning The paper introduces a training-free contrastive decoding method where a small model guides a larger model toward shorter reasoning chains. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates reasoning efficiency rather than developing an LLM evaluation framework. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an execution framework rather than a general LLM evaluation benchmark. | |
| RobustCNN-4: An Efficient Deep Learning Framework for Deepfake Detection This paper evaluates an ensemble deep learning framework of two convolutional networks for robust and efficient deepfake video detection. Peer-reviewedOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via ieee · 14 Oct, 06:18 | Not relevant Proposes a deepfake detection model rather than an LLM evaluation methodology or benchmark. | |
| SURD-BENCH: BENCHMARKING UNIFIED MULTI Summary pending (next check). Could not confirmPreprint | LLM evaluation and benchmarks via openreview · 14 Oct, 06:18 | Waiting for check | |
| MEMTRACE: TRACING AND ATTRIBUTING ERRORS Summary pending (next check). Could not confirmPreprint | LLM evaluation and benchmarks via openreview · 14 Oct, 06:18 | Waiting for check | |
| REASONOPS: OPERATOR SEGMENTATION FOR LLM ... Summary pending (next check). Could not confirmPreprint | LLM evaluation and benchmarks via openreview · 14 Oct, 06:18 | Waiting for check | |
| FROM FACT OVERWRITING TO CAUSAL EDITING Summary pending (next check). Could not confirmPreprint | LLM evaluation and benchmarks via openreview · 14 Oct, 06:18 | Waiting for check |
2,701–2,760 of 3,345← PreviousNext →