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 |
|---|---|---|---|
| Recursive Self-Improvement in Unified Multimodal Models The authors introduce a training loop where multimodal models use text and visual abilities alongside program execution to generate their own training data. PreprintOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via blog · 18 Oct, 06:22 | Not relevant Proposes a training framework rather than evaluating LLM benchmarks or judge mechanisms. | |
| Graphectory Viewer: A Tool for Process-Centric Analysis of Agentic Software Trajectories The paper introduces an interactive web-based tool for analyzing and visualizing execution trajectories of software engineering agents. Peer-reviewedOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via acm · 18 Oct, 06:18 | Not relevant A trajectory inspection tool rather than a benchmark or evaluation method. | |
| Where Do AI Coding Agents Fail? An Empirical Study of Failed Agentic Pull Requests in GitHub The paper analyzes 33,000 GitHub pull requests to examine why AI coding agent contributions succeed or fail. Peer-reviewedOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via acm · 18 Oct, 06:18 | Not relevant Empirical study of pull requests, not benchmark creation or evaluation. | |
| Zero-RAG: Towards Retrieval-Augmented Generation with Zero Redundant Knowledge The paper introduces a method to reduce redundant external knowledge retrieval by estimating what a language model already knows. Peer-reviewedOlder paper (2025) Older paper (2025) — discussed this week | LLM evaluation and benchmarks via blog · 17 Oct, 06:24 | Not relevant Focuses on corpus pruning and routing methods rather than general evaluation frameworks. | |
| A Survey of Secure Retrieval-Augmented Generation This survey paper reviews attacks, defenses, and evaluation methods across 135 studies on secure retrieval-augmented generation systems. PreprintOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via blog · 17 Oct, 06:24 | Not relevant Primary focus is security taxonomy and threats, not general evaluation methodologies or benchmarks. | |
| Measurement-Based Evaluation of Inhomogeneity Effects on PE Path-Loss Prediction The paper examines how along-path evaporation duct variations affect parabolic equation path-loss predictions using maritime radio measurements. Peer-reviewedOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via ieee · 17 Oct, 06:19 | Not relevant Studies maritime radio path-loss prediction, not language model benchmarking. | |
| Concept Map Mining: A Definition and a Framework for Its Evaluation The paper provides a formal definition and an evaluation framework for concept map mining techniques. Peer-reviewedOlder paper (2008) Older paper (2008) — discussed this week | LLM evaluation and benchmarks via ieee · 17 Oct, 06:19 | Not relevant Abstract lacks detail and focuses on concept map mining. | |
| Electrical Impedance Spectroscopy Of Single Particles By Ac Nanopore Method-Toward Evaluation Of Dielectric Properties Of Single Nanoparticles- The paper studies electrical impedance spectroscopy using an AC nanopore method to evaluate the dielectric properties of single nanoparticles. Peer-reviewedOlder paper (2023) Older paper (2023) — discussed this week | LLM evaluation and benchmarks via ieee · 17 Oct, 06:19 | Not relevant Addresses electrical impedance spectroscopy, not language model evaluation benchmarks. | |
| Rederived Closed-Form Continuous-Time Neural Networks The paper introduces an approximation method using Lagrange interpolation to derive closed-form solutions for continuous-time neural networks. Peer-reviewedOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via ieee · 17 Oct, 06:19 | Not relevant Focuses on continuous-time neural networks, not LLM benchmark evaluation. | |
| High Resolution ADC for Ultrasound Color Doppler Imaging Based on MASH Sigma-Delta Modulator The paper presents a high-resolution analog-to-digital converter architecture using a sigma-delta modulator for ultrasound color Doppler imaging. Peer-reviewedOlder paper (2020) Older paper (2020) — discussed this week | LLM evaluation and benchmarks via ieee · 17 Oct, 06:19 | Not relevant Discusses medical ultrasound hardware converters, not language model evaluation. | |
| SAJD: Self-Adaptive Jamming Attack Detection in AI/ML Integrated 5G O-RAN Networks This paper introduces an adaptive framework to detect radio jamming attacks within machine learning integrated 5G Open RAN architectures. Peer-reviewedOlder paper (2025) Older paper (2025) — discussed this week | LLM evaluation and benchmarks via ieee · 17 Oct, 06:19 | Not relevant Focuses on wireless signal telemetry rather than LLM benchmarks. | |
| Oriented Tiny Object Detection: A Dataset, Benchmark, and Dynamic Unbiased Learning This paper presents a dataset, benchmark, and coarse-to-fine learning scheme to detect tiny oriented objects in imagery. Peer-reviewedOlder paper (2024) Older paper (2024) — discussed this week | LLM evaluation and benchmarks via ieee · 17 Oct, 06:19 | Not relevant Evaluates object detection vision models rather than language models. | |
| Reimagining Peer Review Process Through Multi-Agent Mechanism Design This position paper proposes multi-agent reinforcement learning and mechanism design to reform academic peer review processes. Peer-reviewedOlder paper (2026) Older paper (2026) — discussed this week | LLM evaluation and benchmarks via acm · 17 Oct, 06:19 | Not relevant Discusses scientific peer review processes rather than language model evaluation. | |
| A Survey of AI for Materials Science: Foundation Models, LLM Agents, Datasets, and Tools This paper surveys foundation models, agent systems, datasets, and computational tools applied to materials science discovery. Peer-reviewedOlder paper (2025) Older paper (2025) — discussed this week | LLM evaluation and benchmarks via acm · 17 Oct, 06:19 | Not relevant Surveys materials science AI applications rather than general LLM evaluation methods. | |
| Global-in-time Spectral Duhamel-Picard Method for Post-Exposure Bake Simulation and Physics-Informed Neural Operators for EUV Lithography This paper presents a spectral Duhamel-Picard method and physics-informed neural operator for simulating post-exposure bake processes in semiconductor manufacturing. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Physical simulation for lithography, not LLM evaluation or benchmarking. | |
| A polylogarithmic bound for list replicability This paper presents a short, simplified proof establishing a polylogarithmic upper bound for list replicability across binary concept classes. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Pure theoretical learning theory note, not LLM evaluation or benchmarking. | |
| Latent Annealed Posterior Sampling with a Variance-Preserving Bridge for Sparsely Observed Random Fields This paper introduces a latent annealed posterior sampling framework with a variance-preserving bridge for inferring random fields from sparse observations. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Addresses posterior sampling in spatial statistics, not LLM benchmarking. | |
| Graph Additive Models disentangling feature, topology, and interactions This paper proposes an interpretable graph neural network architecture that separates node features, topological motifs, and feature-topology interactions into additive terms. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Proposes a graph neural additive network, not LLM benchmarking or evaluation. | |
| On the Stability and Convergence of Flow Map Training This paper provides theoretical analysis explaining why stop gradients and adaptive weighting are necessary for stable training of generative flow maps. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Analyzes generative flow maps, not LLM benchmarking or evaluation methods. | |
| Spectral Indistinguishability and Learning-Curve Identifiability in Quadratic Networks This paper theoretically analyzes what spectral measurements and learning curves reveal about internal representations in quadratic neural networks. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Theoretical learning theory paper on quadratic networks, not LLM benchmark evaluation. | |
| Diminishing Returns of Data in Sparse Autoencoders: Functional Behavior does not Guarantee Feature Reproducibility This paper examines how training data size affects the functional performance and feature reproducibility of sparse autoencoders on language models. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Evaluates sparse autoencoder training properties rather than LLM benchmarks or evaluation methods. | |
| Causal Syntax Circuits in CodeLMs: Budget and Probe Decide What Looks Localised This paper investigates internal model circuits and prompt properties that drive code language models to default to Python code output. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on mechanistic interpretability of code models, not general benchmark evaluation. | |
| RoLaQ: Joint Rotation and Range Calibration for 2-Bit KV Caches in Multi-Head Latent Attention This paper presents a calibration method to compress multi-head latent attention key-value caches to two bits for efficient long-context language model inference. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Evaluates 2-bit quantization schemes, not general LLM evaluation benchmarks. | |
| On-Policy Residual Learning for Diffusion Language Models This paper introduces an on-policy residual learning framework with an entropy-contraction objective to improve generation speed and quality in diffusion language models. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Studies diffusion language model training and serving, not evaluation benchmarks. | |
| Quantile-Guided Diffusion for Probabilistic Time-Series Forecasting This paper proposes a quantile-guided residual control framework to improve uncertainty representation and accuracy in diffusion-based time-series forecasting models. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Evaluates time-series diffusion forecasting methods, not LLM evaluation or benchmarks. | |
| Vertical Sliced Wasserstein Distances for Meta-Measures This paper introduces a scalable vertical sliced Wasserstein distance to measure distances between probability distributions whose individual elements are themselves distributions. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Proposes a mathematical optimal transport distance, not LLM evaluation benchmarks. | |
| Ergodic topology sampling toward the continuum with Stochastic Normalizing Flows This paper uses stochastic normalizing flows with open boundary conditions to mitigate topological freezing in lattice gauge theory simulations. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Applies normalizing flows to lattice gauge theory, not LLM evaluation. | |
| Semantic Alignment of EEG and Text: A Contrastive Learning Framework for Decoding Imagined Speech This paper aligns electroencephalography signals with text representations in a shared embedding space to improve the decoding of imagined speech intent. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Evaluates imagined speech decoding from EEG, not LLM evaluation benchmarks. | |
| Selection-Safe Conformal Early Exit: Distribution-Free Coverage under Adaptive Depth This paper presents a pathwise conformal prediction method that guarantees valid coverage across adaptive early-exit stopping policies without multi-exit score inflation. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on conformal prediction theory for early exits, not LLM evaluation benchmarks. | |
| Sliced Wasserstein Sketches with Error Certificates This paper develops sliced Wasserstein sketches with computable error certificates for compact representation and fast comparison of probability measures. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Studies mathematical distribution sketching, not LLM benchmarks or evaluation methods. | |
| EEG-Based Prediction of Music Enjoyment Using Deep Learning with Explainable Temporal Analysis This paper evaluates deep learning models and temporal interpretability techniques to predict continuous music enjoyment ratings from electroencephalography signals. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Presents an EEG music enjoyment regression benchmark, not LLM evaluation. | |
| Optimal Explicit and Implicit Regularization Strengths in High-Dimensional Continual Linear Regression This paper derives closed-form generalization loss expressions for high-dimensional continual linear regression, characterizing optimal step sizes and equivalent regularization strengths. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Analyzes theoretical continual linear regression, not LLM evaluation benchmarks. | |
| O-RAN-Native Jamming Detection Under Network Heterogeneity This paper develops a statistical ensemble method to detect radio-frequency jamming in heterogeneous open radio access networks using standard telemetry. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Evaluates RF jamming detection methods, not language model evaluation benchmarks. | |
| Feasibility Ceilings Before Learning: Sizing and Timing Bounds for Drone-Mesh Interception This paper establishes analytical geometric bounds on coverage probability and timing uncertainty for multi-agent drone meshes intercepting approaching objects. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Mathematical coverage bounds for robotics, not language model evaluation. | |
| Linear Independence of Polynomial Compositions with Bounded Input Degree This paper resolves mathematical conjectures on polynomial compositions, proving the identifiability of neural networks using general polynomial activation functions of sufficient degree. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Pure algebraic geometry proof regarding neural identifiability, not LLM evaluation. | |
| Task-Phase-Conditioned Expert Routing Is Not a Reliable Proxy for Quantization Sensitivity in MoE FFN Tensors This paper shows that routing statistics across task phases fail to provide an effective guide for mixed-precision quantization in mixture-of-experts models. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Tests routing-based quantization sensitivity heuristics rather than broad LLM evaluation methodology. | |
| A Phenomenology of Neural Geometry in Vision-Language-Action Models This paper investigates internal geometric representations of task-relevant physical concepts within vision-language-action policies for robotic manipulation. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Analyzes internal geometric representations in VLA models, not general LLM evaluation benchmarks. | |
| Analog RF Computing for Integrated Data and AI Flows Over Wireless Networks This paper investigates using analog radio frequency computing over wireless networks to deliver low-power neural network inference to resource-constrained edge devices. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on wireless RF computing, not LLM evaluation benchmarks. | |
| Denoising Score Matching in Diffusion Models: The Effect of Rank Constraints on Generalization This paper analyzes the effect of rank constraints on generalization and test error in diffusion score matching networks. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Theoretical generalization bounds for diffusion score matching, not LLM evaluation. | |
| Complete or Sparse: A Tale of Two Identifiabilities This paper proves conditions under which supervised neural networks can uniquely identify and recover true underlying data latents in generative and discriminative settings. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Theoretical mechanistic interpretability study, not LLM benchmark or evaluation methodology. | |
| Time-partitioning Gaussian Process Mixture Model This paper introduces a Bayesian nonparametric Gaussian process mixture model using an adaptive Pitman-Yor prior to detect persistent regimes in time series. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Proposes a Bayesian time-series model rather than LLM evaluation. | |
| SNOAR: Self-training with Noise filtering, Observation adding, and Remixing for Speech Denoising Without Clean References This paper proposes a self-training speech denoising framework that corrects teacher predictions via noise filtering and observation mixing without clean references. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Evaluates speech denoising algorithms rather than LLM benchmarks or evaluation methods. | |
| Long-Context Drafting Does Not Require Long-Context Drafters This paper demonstrates that frozen short-context drafting models can effectively serve ultra-long context speculative decoding without fine-tuning on long sequences. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on speculative decoding speedups, not benchmark or evaluation methodology. | |
| Clinically Interpretable Biological Age Estimation Using CatBoost and SHAP with Continuous and Categorical Features from NHANES This paper uses CatBoost and an automated SHAP analysis pipeline on clinical data to estimate biological age with interpretable biomarker insights. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Clinical tabular regression interpretability, not LLM evaluation or benchmarks. | |
| Machine-learning-based neutrino-event generation constrained by experimental measurements This paper introduces a generative surrogate model for neutrino-nucleus scattering calibrated directly to measured experimental cross sections without event-level data. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant High-energy physics simulation paper, not LLM evaluation or benchmarks. | |
| Do Emergent Misalignment Models Share a Circuit? This paper shows that fine-tuned language models exhibiting emergent misalignment across different harmful domains share specific attention heads responsible for the behavior. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Mechanistic interpretability study, not LLM benchmark development or evaluation. | |
| Learning Warm Starts for the Simplex Method from Measured Solver Cost This paper analyzes why machine-learning-predicted warm starts for the simplex method reduce iterations much more than actual solve time. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Studies linear programming solver speedups, not LLM benchmark evaluation. | |
| Flexible Doubly Robust Proximal Causal Inference for Censored Outcomes This paper presents a flexible doubly robust proximal learning framework to estimate causal effects on right-censored time-to-event outcomes under unmeasured confounding. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Biostatistical causal inference framework, not LLM evaluation or benchmarks. | |
| Stochastic Gradient Descent: A Nonlinear Time Series Perspective The paper analyzes the convergence and asymptotic properties of stochastic gradient descent with constant step size using nonlinear time series theory. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Theoretical analysis of stochastic optimization, not LLM evaluation benchmarks. | |
| Sign Betting for Misspecified Regression: Variance Adaptivity and Optimal Approximation The paper analyzes the theoretical statistical properties and oracle inequalities of sign-betting regression under model misspecification. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Theoretical regression analysis, not LLM evaluation or benchmark design. | |
| Recursive Offline Distillation via Entropy Gain Sampling and Answer Mixing The paper presents an offline distillation method combining entropy-difference sampling and answer mixing to transfer reasoning abilities into smaller language models. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on distillation training methods, not LLM evaluation frameworks. | |
| Planning to Learn: Real-Time Robust Adaptive Control on the GPU with Guaranteed Future Learning The paper presents a GPU-accelerated robust adaptive control framework that accounts for guaranteed future parameter learning during real-time robotic trajectory planning. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Robotics control algorithm, not LLM evaluation or benchmark methodology. | |
| Ranked Right, Sized Wrong: Ensemble Spread Fails Silently at Streamflow Extremes The paper demonstrates that ensemble spread in streamflow forecasting models becomes significantly under-dispersed and underestimates error during extreme flood events. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on hydrological forecasting uncertainty, not LLM benchmark evaluation. | |
| Cross-subject decoding of imagined speech from scalp EEG: a leakage-controlled evaluation on three public corpora The paper evaluates cross-subject transfer and accuracy limits of decoding imagined speech from scalp electroencephalography signals across three public datasets. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Studies EEG brain decoding transfer, not language model evaluation. | |
| MIDI-REPA: Improving Text-to-Music Generation with Symbolic Representation Alignment The paper aligns intermediate representations of text-to-music models with symbolic music transcriptions to improve generation quality efficiently. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on text-to-music representation learning, not LLM evaluation. | |
| Two-Stage Score Estimation for Group-Specific Diffusion Models The paper establishes theoretical convergence rates for group-specific diffusion models that learn a shared score alongside a target-specific offset. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Studies theoretical density estimation, not language model benchmark evaluation. | |
| The Blume-Capel Model as a $φ^6$ Field Theory: A Monte Carlo and Flow Model Study The paper evaluates continuous spin Blume-Capel models using Monte Carlo simulations and normalizing flows to accurately map phase diagrams. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Statistical physics sampling paper, not LLM evaluation or benchmarks. | |
| Halve-Then-Greedy: Adaptive Sampling for Posterior Success with Near-Optimal Instance Dependence The paper presents an adaptive bandit sampling algorithm designed to verify that an option exceeds a performance threshold within a fixed budget. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on multi-armed bandit sampling theory, not LLM benchmarks. | |
| Imperfect Is Perfect: Learned Column Prefilling for Branch-Price-and-Cut The paper uses neural networks to predict initial columns and dual centers, speeding up branch-price-and-cut solvers for vehicle routing problems. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Addresses operations research optimization, not LLM evaluation methods. | |
| CausalEvo: Causal Skill Self-Evolution for Health LLM Agents The paper introduces a causal evaluation method for healthcare language model agents to iteratively select and retain useful task skills. Preprint | LLM evaluation and benchmarks via arxiv · 17 Oct, 06:18 | Not relevant Focuses on agent skill evolution, not benchmark methodology or contamination. |
1–60 of 2,721Next →