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 |
|---|---|---|---|
| Relational Learning as World Modeling The paper presents TabWM, a tabular world model that infers system dynamics from past executions to predict database edits from actions. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents tabular world models, not LLM evaluation. | |
| Geospatial Vector Geometry as a Native Spatial Modality for Vision-Language Models The paper proposes GeoVec-VLM, integrating geospatial vector geometry as a native input modality to improve spatial reasoning in vision-language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a multimodal architecture, not benchmark evaluation. | |
| Game-Creation-Agents: Autonomous Game Production The paper introduces an autonomous multi-agent framework that translates natural language prompts into playable browser games with automated verification. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on multi-agent game production, not benchmarks. | |
| PrivacyGuard: Source-Linked Sanitization and Locally Resolvable PII References for Cloud GUI Agents The paper presents PrivacyGuard, a framework using source-linked sanitization and local references to prevent personal data leakage in cloud GUI agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a privacy defense system, not evaluation methodology. | |
| LIBERO-Industry: Benchmarking Robotic Manipulation under Industrial Process Constraints The paper introduces LIBERO-Industry, a benchmark evaluating robotic manipulation policies under industrial process constraints such as moving material flows and tolerances. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Benchmark covers robot manipulation policies, not language models. | |
| InfoVis: Unified Slide Image Generation and Editing The paper introduces InfoVis, a 9B multimodal model along with a benchmark and training dataset for slide image generation and editing. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Slide benchmark is secondary to generating multimodal presentation slides. | |
| Logit Leakage through Floating-Point Bias Addition The paper reveals a numerical side channel in floating-point bias addition that enables recovering full logit vectors from API queries. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on numerical side-channel attacks, not benchmarking. | |
| Growing to See: Efficient Pretraining of Self-Supervised Vision Transformers The paper investigates progressive depth growth during self-supervised vision transformer pretraining to improve downstream accuracy while reducing computation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies layer growth during training, not LLM evaluation. | |
| DTA-GT: Direction- and Topology-Aware Graph Transformer for Neural Network Representation Learning The paper introduces a direction- and topology-aware graph transformer to predict neural network architecture attributes like accuracy and latency. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on NAS accuracy prediction, not LLM benchmarks. | |
| Unsupervised Data Drift Monitoring for Deployed Deepfake Speech Detectors The paper proposes an unsupervised drift monitor using internal embedding distances to flag unfamiliar inputs in deployed deepfake speech detectors. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on run-time drift monitoring, not LLM benchmarks. | |
| A Scaled-Up Empirical Study of Syntactic Language Models The paper presents an empirical study evaluating syntactic language models at scales up to 1B parameters across linearizations and masks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes syntactic architecture scaling, not evaluation methodologies. | |
| Few-for-Many Best-Arm Identification in Multi-Objective Bandits The paper presents multi-objective bandit algorithms to identify a small set of arms that minimizes degradation across multiple objectives. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on multi-objective bandits, not LLM evaluation. | |
| Towards Cost-Efficient Video Dataset Construction with Large Language Models The paper presents an LLM-based metadata pre-screening method to filter candidate videos before downloading, reducing data curation costs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes dataset curation filtering, not LLM benchmark evaluation. | |
| Davinci-Tool: Scaling Agentic Tool-Use Supervision with Verifiable Executable Environments The paper introduces a 30B tool-use model and an execution-grounded dataset to scale agentic tool supervision with verifiable environments. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focus is on dataset synthesis and agent training. | |
| CLIVO: Differentiable Cluster-Based Instruction Mixing with Learned Validation Objectives The paper proposes CLIVO, a target-aware method combining semantic clustering and learned validation objectives for differentiable multimodal instruction data mixing. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a data selection method rather than an evaluation benchmark. | |
| Acceptable Misalignment The paper argues that full robotic value alignment is impossible and examines how to manage acceptable misalignment in autonomous systems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Theoretical paper on misalignment risks rather than an empirical benchmark. | |
| Transparency as Answerability: Role-Dependent Information Flow Modeling The paper formalizes transparency as role-dependent answerability using information flow networks to evaluate AI-supported decision systems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Formulates an analytical framework for transparency rather than a model benchmark. | |
| Observe, Share, or Predict? Disentangling Partner Information in Embodied Multi-Agent Collaboration The paper investigates how observing partner states, sharing knowledge, and predicting partner actions affect coordination in embodied multi-agent teams. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes agent coordination strategies rather than proposing an LLM benchmark. | |
| Neural Bayesian Networks for Modular and Scalable Decision-Making Under Uncertainty The paper proposes Neural Bayesian Networks, replacing classical probability tables with differentiable tensor operations for scalable probabilistic reasoning on GPUs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a neural probabilistic framework rather than an LLM benchmark. | |
| Transparent Information Flow Modeling for Role-Aware AI Act Assessment The paper applies an information-flow modeling methodology to assess public-sector AI recruitment tools under European Union AI Act requirements. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a compliance assessment methodology rather than an LLM evaluation benchmark. | |
| WeFair: Fidelity-Aware Alignment from Human Feedback for Personalized Query Rewriting in Short-Video Search The paper proposes an alignment framework using human feedback signals to personalize search query rewriting without causing intent drift. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an alignment framework rather than an evaluation benchmark. | |
| KNN Implementation Details Can Dramatically Change Performance: An Example from Cover Trees The paper analyzes how Cover Tree nearest-neighbor implementation details and query modes materially change runtime performance across datasets. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes algorithmic implementation details rather than LLM benchmark design. | |
| Social Priors: Descriptions of AI Users Shape How Language Models Trust and Cooperate The paper investigates how fine-tuning language models on descriptions of user motivations influences their downstream trust and cooperative behavior. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies behavioral shifts from finetuning rather than introducing evaluation benchmarks. | |
| Proactive Physical AI through Continuous Observation and Timely Intervention The paper proposes a system for physical AI that continuously monitors environments to identify intervention needs and execute timely actions. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on proactive robotics architecture rather than LLM benchmarking methods. | |
| MADDIE: Multi-Agent Drug Discovery Engine The paper introduces a multi-agent framework comprising an orchestrator and four specialized agents to automate iterative drug discovery workflows. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a multi-agent application system, not an evaluation benchmark. | |
| The Value of Being Wrong: Self-Mined Visual In-Context Learning from Errors The paper introduces an in-context learning method that selects demonstration examples by mining a vision-language model's own past prediction mistakes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an in-context learning method rather than an evaluation benchmark. | |
| WeGR: Unified Generative Retrieval for Industrial Search Systems in WeChat The paper introduces an industrial generative retrieval framework featuring collaborative tokenization and preference optimization for WeChat search systems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a generative retrieval method rather than an evaluation benchmark. | |
| PRAISE: A Reproducible Workflow for Non-Programmer Scientists to Implement Published AI Models The paper presents a structured step-by-step workflow designed to help non-programmer scientists implement and validate published machine learning models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an implementation guide for scientists, not an LLM benchmark. | |
| From Suboptimal to Optimal: One-Step Latent Flow Policies for Offline Reinforcement Learning The paper proposes an offline reinforcement learning method that optimizes generative decision chunks in latent space using conditional velocity fields. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a reinforcement learning algorithm rather than an evaluation benchmark. | |
| 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. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a model architecture and training method, not an evaluation framework. | |
| TabGift: A Context-Efficient Foundation Model for Tabular Data Generation The paper proposes a context-efficient foundation model for generating realistic tabular data from limited context observations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a tabular foundation model rather than an evaluation benchmark. | |
| Governing Copyright Responsibly in the Era of Generative AI The paper analyzes forty legal instruments to examine how copyright governance protects scholarly authors whose works train generative AI models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Examines copyright governance frameworks rather than model performance benchmarks. | |
| MAGNET: Manifold-Aware Graph Diffusion Network for Connectome Generation The paper proposes a manifold-aware graph diffusion network to generate valid functional brain connectomes while capturing fine-grained geometric structure. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a generative architecture rather than general model evaluation methods. | |
| Understanding Personalization Gains in Whisper The paper studies low-rank adaptation in Whisper to evaluate whether transcription gains require group-specific adapters rather than pooled fine-tuning. Peer-reviewed | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies LoRA personalization in speech rather than general LLM evaluation benchmarks. | |
| Benchmarking Quantum Kernel SVMs for Credit Card Fraud Detection: Performance Realities Under NISQ Constraints The paper benchmarks three quantum kernel SVM encodings against classical models for imbalanced credit card fraud detection under NISQ constraints. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Benchmarking classical vs quantum ML rather than LLM evaluation or benchmarks. | |
| SPLICE: Sparse Physics Library Identification and Compositional Expansion Introduces a framework that discovers and separates physical neural operators from unlabeled trajectory data containing multiple mechanisms. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an operator discovery algorithm rather than model evaluation benchmarks. | |
| Neos: Learning to Interleave Search and Visual Reasoning Develops a multimodal search framework that learns to turn intermediate visual discoveries into guidance for subsequent search actions. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an agent search framework rather than a model evaluation benchmark. | |
| PatientWM: Towards Generalizable Clinical World Models via Unified Patient States Proposes a clinical world-model architecture using unified patient states and ordered action histories to anticipate treatment outcomes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a domain modeling architecture rather than a model evaluation benchmark. | |
| Gradient Routing Localizes and Removes Unintended Behaviors in RL Routes gradients from imperfect monitor flags into sacrificial parameters to mitigate reward hacking during reinforcement learning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an optimization technique rather than an evaluation benchmark. | |
| Superintelligence and the Triple Risk of an AI Race Argues that competitive national races toward superintelligent artificial intelligence create risks of conflict and catastrophic loss of control. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant A strategic policy discussion rather than model benchmarking or testing. | |
| What the Machine Invented: Mining New Mathematical Instruments from OpenAI's October 2026 Release of AI-Generated Mathematics Extracts and categorizes mathematical proof mechanisms and lemmas from an open release of synthetic mathematical manuscripts. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes AI-generated mathematical manuscripts rather than evaluating model benchmarks. | |
| AIPOCH Open-Science: An Open-Source AI Research Workbench with Reproducible .science Packages Presents an open-source research workbench and portable packaging format to preserve relationships between computations, artifacts, and assessments. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a reproducible workbench rather than an LLM evaluation benchmark. | |
| Calibrated Typed Decisions for Entity, Relation, and Event Extraction Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates on benchmarks but primarily introduces a new extraction model architecture. | |
| D-World: An Explicit World-Language-Action Model with Recursive Self-Improvement under Dual Control Designs an explicit digital-twin world model where robotic agents inspect, simulate, and calibrate physical parameters before acting. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a model framework rather than evaluating benchmarks or testing methodologies. | |
| D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Model conversion method, not an evaluation benchmark or audit. | |
| Who is the 'I' in AI? Establishing Decentralized Identity and Accountability for Autonomous Agents Proposes a decentralized identity framework using self-sovereign identity principles to establish verification and accountability for autonomous agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Identity specification proposal, not an evaluation benchmark. | |
| EvoQuantLab: Training-Free Dual-Loop Evolution for Auditable Quantitative Research Combines genetic programming and diagnostic evaluation loops to discover quantitative financial trading factors without backtest overfitting. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a factor discovery system, not an LLM benchmark. | |
| Force-Modulated Dexterity: Unified Compliance for Contact-Rich Manipulation Presents a compliant teleoperation control framework that reduces excessive contact forces during dexterous manipulation data collection. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a robotic control technique, not LLM evaluation. | |
| Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Study on odor perception representations, not LLM benchmark evaluation. | |
| Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery Discovers robot capabilities using a generative flow network that creates symbolic task skeletons grounded in physical robot affordances. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a task generation framework, not an evaluation benchmark. | |
| FreeSolo: Linear-time Sparse Attention with Zero-Shot Length Extrapolation Introduces a native sparse attention mechanism that combines sliding windows and keypoint selection for linear-time context extrapolation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on architectural design, not benchmark creation or evaluation. | |
| Size Generalization for Diffusion Routing Solvers Uses reinforcement learning on large routing instances to help diffusion-based solvers generalize beyond their training instance sizes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Covers combinatorial optimization algorithms, not LLM evaluation. | |
| Deleting Code Without Losing Knowledge: Regenerative Software for Python Libraries Applies regenerative software primitives so programs can be deleted and rebuilt by agents without losing architectural intent and knowledge. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Describes a software engineering methodology, not benchmark evaluation. | |
| AutoMARS: Automatic and Scalable Multi-Agent Large Language Model Framework for Supply Chain Risk Assessment The paper proposes a multi-agent framework to coordinate analytical tools and evaluate multi-factor supply chain risks across diverse locations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on multi-agent application design, not benchmark development. | |
| Compute-optimal data scaling for neural surrogates via multi-fidelity training The paper demonstrates multi-fidelity training to optimize data generation compute and error scaling for neural PDE surrogates. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Covers scientific simulator scaling, not LLM evaluation benchmarks. | |
| From Conceptual EneMonIE to an Executable Backend: A Modular Architecture for Energy Information Extractionand Grounded Question Answering The paper describes a modular software backend architecture for extracting energy domain information and answering grounded questions from web sources. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Software system implementation, not evaluation benchmarks or methodology. | |
| RL Post-Training Diffusion Language Models Along Their Own Generation Paths The paper presents a reinforcement learning post-training method that scores intermediate generation trajectories in diffusion language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an RL training algorithm, not an evaluation benchmark. | |
| Measuring Learned Equivariance: An Information Theoretic Approach The paper proposes information-theoretic tests to measure learned equivariance in intermediate neural network representations without knowing explicit group transformations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Deals with representation invariance metrics rather than LLM benchmarks. | |
| Reasoning for Propaganda Detection via Label-Guided CoT Optimisation The paper presents a reasoning-conditioned framework that refines chain-of-thought explanations to improve propaganda detection without explanation annotations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a classification framework rather than benchmark evaluations. | |
| Phase-Conditioned Affordance Fields for SE(3)-Equivariant Visuomotor Policies The paper introduces phase-conditioned affordance fields to improve equivariant visuomotor policy learning for robotic manipulation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a robotic policy method rather than LLM benchmarks. |
2,461–2,520 of 3,345← PreviousNext →