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 |
|---|---|---|---|
| Co-DSL: Human-LLM Co-design of Reusable Spatial Concepts for Layout Generation The paper presents Co-DSL, enabling users and language models to co-design reusable spatial concepts to guide layout generation solvers. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates spatial layout generation, not LLM benchmark methodologies. | |
| OntoAR-HAR: Ontology-Guided Agentic RAG with Personalized Memory for Human Activity Recognition The paper presents an ontology-guided retrieval framework combining population and personalized memory for wearable sensor human activity recognition. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates activity recognition accuracy rather than benchmarking methodology. | |
| HGA-Quant: Heterogeneity-Aware Quantization for AutoRegressive Visual Generation Models The paper introduces a post-training quantization framework for autoregressive visual generation models to retain image quality under low precision. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates quantization performance, not general evaluation methodologies. | |
| CHEAR: Confusion-Hierarchy Evidence-Aware Reasoning for Cost-Efficient Human Activity Recognition The paper introduces CHEAR, an activity recognition framework that pre-organizes confusion hierarchies offline to reduce inference costs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a reasoning framework for sensor data, not an LLM benchmark. | |
| WisdomNet: trustable machine learning toward error-free classification The paper proposes an artificial neural network framework designed to reject uncertain predictions to prevent misclassification errors. Peer-reviewed | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an abstention mechanism for neural classifiers, not an LLM benchmark. | |
| Accuracy and Uncertainty in Machine Deception Detection (Research White Paper) The paper proposes a multi-layer detector ensemble to identify deceptive model outputs while calibrating score uncertainty and false-positive rates. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Develops a detector for deception rather than a broader LLM evaluation framework. | |
| SuperAI: Merging Capabilities Across Heterogeneous Foundation Models The paper introduces SuperAI, a framework to merge complementary capabilities across heterogeneous foundation models using low-dimensional subspace alignment. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes model fusion rather than LLM benchmark analysis. | |
| Anchor-and-Harvest Curriculum: A Fully Unmasked Non-LLM Curriculum for Craftax The paper introduces a rule-based curriculum training system that improves reinforcement learning performance on Craftax without language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a curriculum training method for RL rather than LLM benchmarking. | |
| ZOIF: Zeroth-Order Influence Function Estimation for LLM Data Selection Using Only Forward Passes The paper introduces a zeroth-order influence function framework to enable memory-efficient training data selection using only forward passes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a training data attribution technique, not an evaluation benchmark. | |
| MultiFlow: Vision-Driven Multimodal fMRI Encoding via Stimulus-Informed Source Flow Matching The paper introduces MultiFlow, a generative framework that models distributions of fMRI neural responses using stimulus-informed flow matching. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Addresses computational neuroscience encoding benchmarks rather than LLM evaluation. | |
| StimFlow: Synthesizing Brain Activity from Images via Stimulus-conditioned Flow Matching The paper presents StimFlow, using conditional flow matching to synthesize fMRI brain activity directly in voxel space from visual stimuli. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates neural response synthesis, not general LLM benchmarking. | |
| Across Viewpoints and Time: Trajectory-Generalizable Physical Attacks on World Models and Video Understanding Models The paper develops an attack framework to generate physical adversarial textures that deceive video models across varying trajectories and viewpoints. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an adversarial attack method rather than standard evaluation methodologies. | |
| Gugak-VocalSet: A Multi-Genre Studio-Quality Dataset of Korean Traditional Singing with Sigimsae Annotations The paper releases a studio-quality audio dataset of Korean traditional singing with dense annotations for vocal expressions and melodic contour. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a traditional Korean music audio dataset, not LLM benchmarks. | |
| Siam-ALS: An Explainable Siamese Deep Learning Framework for Early ALS Detection Using EEG and Eye-Tracking Signals The paper introduces an explainable Siamese deep learning framework combining synchronized EEG and eye-tracking signals for early ALS detection. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates a clinical model, not LLM evaluation frameworks. | |
| Adversarial Detection for Pre-trained Classifiers via Lipschitz-Margin Analysis with Certification The paper proposes certified adversarial detection using Lipschitz-margin analysis to detect perturbations rather than forcing invariant predictions. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on certified adversarial detection methods, not general evaluation benchmarks. | |
| TRACE: Temporal Risk Assessment and Concept Editing for Safer Language Generation The paper introduces TRACE to monitor harmful response risks during generation, edit token activations selectively, and maintain auditable decision traces. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a safety steering framework rather than a general evaluation benchmark. | |
| A Sovereign, Indic-First, Reasoning Model The paper presents a hybrid language model adapted for reasoning, instruction following, and tool use across English and eight Indian languages. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Reports benchmark performance without investigating evaluation methodology. | |
| Fewer Bits, More Tries: Recovering Capability Under Memory Constraints This paper shows that running multiple parallel attempts with a smaller model can match the coding accuracy of much larger models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates on standard benchmarks rather than analyzing benchmark methodology. | |
| In-Place Indic Tokenizer Expansion and Continual Pretraining of Nemotron-3-Nano-30B-A3B This paper details vocabulary expansion and continual pretraining techniques to adapt an open language model across eight Indic languages. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates on benchmarks but does not introduce evaluation methodologies. | |
| Analyzing the Dynamics of WSBM-based Echo State Networks using InputDSA This paper analyzes the computational dynamics of biologically structured modular reservoir networks using a comparative dynamic similarity framework. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies reservoir computing dynamics, not evaluation benchmark methodologies. | |
| E8 Lattice Key Quantization for Low-Bit KV-Cache Compression This paper presents a lattice vector quantization method that compresses attention key-value caches to two bits while preserving model quality. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a compression technique, not a general evaluation benchmark. | |
| A Decade of PhotoMappers: A Longitudinal Study of Crowdsourced Disaster Photos Geolocalization with Vision-Language Models and Geospatial Reasoning This study analyzes a decade of crowdsourced disaster photos to evaluate vision-language models against human geographic reasoning for location estimation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a domain-specific study, not a general evaluation benchmark. | |
| State Representation Shapes Long-Horizon Agent Learning and Transfer: An Empirical Study This empirical study evaluates how compact versus full state representations affect the ability of language model agents to learn long-horizon workflows. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes state representations in agents, not general evaluation benchmarks. | |
| SpecFit: Not All Draft Tokens Cost the Same in Offloaded MoE Inference This paper proposes a speculative decoding tree selection strategy that accounts for host-to-device transfer costs during offloaded mixture-of-experts inference. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Optimizes inference throughput rather than model evaluation benchmarks. | |
| Pessimistic Latent Task-aware Optimization for Robust Offline Meta-Reinforcement Learning This paper proposes a pessimistic offline meta-reinforcement learning method that generates synthetic off-support latents to improve out-of-distribution task transfer. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an RL optimization algorithm, not an evaluation benchmark. | |
| World Modeling through Spectral Alignment This paper introduces a world-modeling framework that uses kernel similarity to control which physical properties are preserved in learned embeddings. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a representation learning method, not an evaluation benchmark. | |
| Hidden in Plain Sight: Advancing Harmful Content Understanding via Perception-Aware Adversarial Self-Play This paper presents an adversarial self-play framework using multi-perceptual views to improve language model detection of obfuscated harmful text. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an adversarial training method, not a general evaluation benchmark. | |
| BridgeCodec: Mamba Enhanced Neural Audio Codec with Schrödinger Bridge at Low Bitrate This paper proposes an optimal-transport bridge framework to translate audio representations directly between mismatched neural codecs at low bitrates. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an audio translation architecture, not an evaluation benchmark. | |
| When Messages Go Wrong: Understanding Communication Faults in LLM-Based Multi-Agent Systems This paper provides an empirical study and taxonomy of communication faults and error propagation across multi-agent language model systems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Investigates system communication faults rather than LLM benchmarking standards. | |
| Scalable Robust MCTS with Anytime Deterministic Optimality Guarantees This paper develops an online tree search method providing deterministic anytime bounds for robust planning under uncertain transition dynamics. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an optimization algorithm, not an evaluation benchmark. | |
| MemeLogos: Knowledge-Driven Retrieval-Augmented Reasoning for Harmful Meme Detection This paper introduces a retrieval-augmented reasoning framework that extracts recurring patterns to detect harmful multimodal social media memes. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a harmful meme detection method, not a benchmark. | |
| SkillFusion: Towards Reliable Multi-Skill Composition for LLM Agents This paper introduces an offline compiler that detects conflicts and merges multiple agent skills into a single coherent capability. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on skill merging architecture rather than evaluation benchmarks. | |
| When Skills Remain But Access Fails: Factorized Skill Memory for Continual Skill Internalization This paper investigates catastrophic forgetting in continual agent skill learning and proposes a factorized memory architecture to preserve procedural execution. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a skill memory architecture, not an evaluation framework. | |
| The Imperishable Record, Forgiveness, and the Future of Criminal Justice This paper examines how persistent digital records and AI influence individual futures, proposing legal and conceptual limits on the eternal use of criminal data. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Addresses criminal justice ethics, not model evaluation or benchmarks. | |
| Large Decision Models: Repurposing Large Language Models as Single-Pass Decision Makers This paper adapts language models into single-pass decision-makers that output explicit probability distributions over task labels specified at run time. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a classification decision model, not an LLM evaluation methodology. | |
| Data Portability Revisited: Toward the Human-Centric, AI-Driven Data Ecosystems of Tomorrow This article analyzes current legal and technical frameworks for data portability, arguing for a human-centric approach to personal data governance in AI. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Discusses regulatory data frameworks, not LLM benchmark evaluation. | |
| CacheLink: Functional KV Transport for Zero-Prefill Cross-Model Context Handoff This paper introduces a method to translate key-value caches between different language models, enabling context handoffs without reprocessing input tokens. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a cache serving method, not an evaluation benchmark. | |
| SARATHI: Decoupled Data-Free Probing and Adaptive Reconstruction for Structured LLM Pruning This paper presents a structured pruning framework that decouples data-free neuron selection from localized weight reconstruction to reduce model memory requirements. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a pruning technique, not an LLM benchmark. | |
| OceanFlow: Coherent Flow Matching for Probabilistic Ocean Modeling The paper proposes a flow matching framework for probabilistic global ocean modeling to generate calibrated conditional ensembles. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on physical ocean modeling, not LLM evaluation benchmarks. | |
| Improving the Bounds on Approximating Weighted Maximin Share for Chores The paper establishes improved approximation bounds and polynomial-time algorithms for allocating indivisible chores fairly among weighted agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Theoretical chore allocation paper, not an LLM evaluation benchmark. | |
| ELECT: Settling Identity Before Rendering for Scalable Story Visualization The paper proposes a training-free method establishing entity identity before final rendering to maintain visual consistency across long storyboards. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a visual rendering method, not an evaluation benchmark. | |
| Self-Representation Alignment in Diffusion Models for Video Background Music Generation The paper introduces an internal feature alignment loss for diffusion models to improve background music generation from video inputs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a generative audio-visual architecture, not an evaluation benchmark. | |
| Inverse Garment Design with Visual Agents The paper introduces an agentic system that reconstructs multi-layer 3D garment sewing patterns from a single image using simulation feedback. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an inverse design framework, not an evaluation benchmark. | |
| CondMSP: Condition-Informed and Resource-Aware Inorganic Materials Synthesis Planning The paper presents an inorganic materials synthesis framework that incorporates condition prediction as context to guide reaction route planning. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Addresses materials science route planning, not LLM evaluation. | |
| From Stages to States: Unified Iterative Reasoning for MLLMs The paper proposes an iterative reasoning framework that models intermediate multimodal reasoning steps as dynamic state transitions. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a multimodal reasoning technique, not an evaluation framework. | |
| Recall What Is Missing: Composing Evidence from Long-Term Agent Memory The paper presents a training-free memory selection method that progressively composes non-redundant evidence sets for long-term agents. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Evaluates a memory mechanism, not proposing general LLM benchmarks. | |
| Dynamic Conflict Clustering with Lightweight Learning and Strategic Repair for Distributed Constraint Satisfaction The paper introduces a conflict-clustering solver with lightweight learning to coordinate distributed agents solving constraint satisfaction problems. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents a DCSP solver, not an LLM evaluation benchmark. | |
| Knowledge Interaction in Hypernetwork-based Continual Learning: Conflict and Composition The paper examines how hypernetwork adapters resolve knowledge conflicts and compose information during continual learning in language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Analyzes internal knowledge interactions, not evaluation benchmarking. | |
| Rethinking Human Intent-to-CAD: Parametric CAD Model Generation via Cooperative Multi-Task Alignment and Spatial-Aware Reinforcement Learning The paper presents a dataset and framework mapping sketches and text to executable parametric computer-aided design code. Peer-reviewed | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces a CAD dataset/method, not an LLM benchmark. | |
| Radioactive Feedback: Tracing Models Trained by Label-Only LLM Supervision The paper proposes an evaluation technique that embeds behavioral traces into score-only feedback to trace distilled student models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies distillation provenance, not general evaluation methodologies. | |
| FARL: A Fully Automated Pipeline of RL Interface and Solver Design The paper presents an automated framework that jointly designs reinforcement learning observation spaces, reward programs, algorithms, and training configurations. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an automated RL search pipeline, not a benchmark. | |
| TabMosaic: Unified In-Context Diffusion for Tabular Generative Tasks The paper presents an in-context diffusion model for generating, completing, and synthesizing unseen tabular data records jointly. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a tabular generative model, not an evaluation benchmark. | |
| RespFocus: Learning Temporal Evidence for Weakly Supervised Respiratory Sound Classification The paper introduces a weakly supervised learning framework and dataset for identifying acoustic evidence in respiratory sound classification. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Medical sound dataset study, not general LLM evaluation. | |
| SpecForge: Recursive Self-Improvement for Coding Agents via Behavioral Specifications The paper presents a self-improvement method where coding agents refine natural-language behavioral specifications using execution-verified tests. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes an agent improvement framework rather than a testing benchmark. | |
| What DNA Foundation Models Learn Beyond Sequence Composition The paper evaluates whether genomic foundation models offer predictive value over simple sequence frequency baselines across classification benchmarks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies biological sequence representations, not LLM evaluation methods. | |
| History Replays Itself: Unleashing Long-Context Capability in Linear Attention The paper proposes a training-free episodic memory mechanism that extends linear attention language models to long-context tasks. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an architectural memory module rather than an evaluation benchmark. | |
| Uncovering the Generative Landscape of Diffusion Models: A Spectral-Guided Quality-Diversity Diffusion Framework The paper integrates quality-diversity optimization with diffusion models to generate diverse sets of high-performing outputs. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Introduces an optimization algorithm for diffusion, not LLM evaluation. | |
| SkillMarker: A KGW-Based Watermarking Method for Agent Skill Ownership Verification The paper introduces a watermarking method to verify ownership and trace unauthorized copying of agent skill instructions. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on watermarking mechanisms rather than evaluation benchmark methodology. | |
| Hyperbolic Time Series Forecasting with Learnable Decomposition and Manifold Dynamics The paper proposes a hyperbolic time series forecasting framework that maps multiscale frequency hierarchies onto the Poincaré ball. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a hyperbolic forecasting architecture, not an LLM evaluation benchmark. | |
| SurgiCLAW: An Interpretable Atomic-Skill Agent for Precision Robotic Micromanipulation The paper presents a hierarchical agent combining vision-language models for task planning with deterministic control programs for micromanipulation. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a surgical robotic framework rather than general LLM evaluation. |
2,281–2,340 of 3,345← PreviousNext →