Every paper checked — nothing hidden

All papers

Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.

All17,313 In brief24 More (on-topic)3,333 Not relevant13,754 Not assessed87 Older352 Could not confirm93 Blogs with no paper266 All blogs284

Paper	Topic	Relevance
ReAl: Refine neural operators via aligning dynamical flow This paper introduces a flow-matching plug-in module to recover high-frequency physical structures in pretrained neural partial differential equation solvers. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant PDE surrogate modeling technique, not small models or efficient inference.
ReAl: Refine neural operators via aligning dynamical flow This paper introduces a flow-matching plug-in module to recover high-frequency physical structures in pretrained neural partial differential equation solvers. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Neural operator refinement for PDEs, not LLM evaluation benchmarks.
Hypothesis Selection via Weighted Medians This paper provides a fast hypothesis selection algorithm using weighted medians that improves runtime dependence on failure probability. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Pure statistical hypothesis testing theory, not LLM benchmark evaluation.
Hypothesis Selection via Weighted Medians This paper provides a fast hypothesis selection algorithm using weighted medians that improves runtime dependence on failure probability. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Theoretical distribution learning algorithms, not model compression or efficient serving.
Hypothesis Selection via Weighted Medians This paper provides a fast hypothesis selection algorithm using weighted medians that improves runtime dependence on failure probability. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Theoretical statistics on sample complexity, not AI safety or deployment reliability.
From Human Corrections to Learned Segmentations: A Critical Mathematical Review of Information, Learning, and Human Effort This paper presents a mathematical review analyzing when human corrections translate into reusable learning versus interactive refinements in image segmentation. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on interactive image segmentation learning, not AI safety guardrails.
From Human Corrections to Learned Segmentations: A Critical Mathematical Review of Information, Learning, and Human Effort This paper presents a mathematical review analyzing when human corrections translate into reusable learning versus interactive refinements in image segmentation. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Reviews interactive segmentation feedback dynamics, not LLM benchmarking or evaluation.
From Human Corrections to Learned Segmentations: A Critical Mathematical Review of Information, Learning, and Human Effort This paper presents a mathematical review analyzing when human corrections translate into reusable learning versus interactive refinements in image segmentation. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Reviews human feedback in image segmentation, not model efficiency or distillation.
HRTF Upsampling and Personalization via Neural Operators This paper develops a neural operator framework to upsample and personalize continuous head-related transfer function fields from sparse spatial measurements. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Domain-specific spatial audio modeling, not AI safety or reliability.
HRTF Upsampling and Personalization via Neural Operators This paper develops a neural operator framework to upsample and personalize continuous head-related transfer function fields from sparse spatial measurements. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Neural operator for head-related transfer functions, not LLM compression or inference efficiency.
HRTF Upsampling and Personalization via Neural Operators This paper develops a neural operator framework to upsample and personalize continuous head-related transfer function fields from sparse spatial measurements. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on binaural audio upsampling, not LLM benchmarking.
Training Language Models to Fail Gracefully This paper introduces a reinforcement learning approach that trains language models to execute safe fallbacks instead of unreliable guesses. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes a training method for fallback behavior, not an evaluation benchmark.
Training Language Models to Fail Gracefully This paper introduces a reinforcement learning approach that trains language models to execute safe fallbacks instead of unreliable guesses. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Presents an RL training method for fallbacks, not model compression or serving.
Posterior Frequencies and the Cost of Not Fitting: A Correction to a Memorization Lower Bound This paper provides counterexamples to an established memorization lower bound and derives exact replacements using full-sample posterior probabilities. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Statistical learning theory on memorization bounds, not LLM evaluation benchmarks.
Posterior Frequencies and the Cost of Not Fitting: A Correction to a Memorization Lower Bound This paper provides counterexamples to an established memorization lower bound and derives exact replacements using full-sample posterior probabilities. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on mathematical bounds of sample memorization, not AI safety monitoring.
Posterior Frequencies and the Cost of Not Fitting: A Correction to a Memorization Lower Bound This paper provides counterexamples to an established memorization lower bound and derives exact replacements using full-sample posterior probabilities. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Theoretical bound correction for statistical learning, not efficient model architectures.
Hyperspherical Contrastive Representations Unlock Planning and Exploration This paper presents a self-supervised reinforcement learning method that uses hyperspherical representations and spherical interpolation to infer planning waypoints. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on robotic manipulation representations, not LLM benchmarks.
Hyperspherical Contrastive Representations Unlock Planning and Exploration This paper presents a self-supervised reinforcement learning method that uses hyperspherical representations and spherical interpolation to infer planning waypoints. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Proposes contrastive representations for robot planning, not small model optimization.
Hyperspherical Contrastive Representations Unlock Planning and Exploration This paper presents a self-supervised reinforcement learning method that uses hyperspherical representations and spherical interpolation to infer planning waypoints. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Studies robotic exploration planning, not language model safety or guardrails.
Tracking a Small Set of Experts in Online Convex Optimization This paper proves a computational separation between information-theoretic and polynomial-time regret when tracking small expert subsets in online convex optimization. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Theoretical study of expert tracking bounds, not model compression or efficient serving.
Tracking a Small Set of Experts in Online Convex Optimization This paper proves a computational separation between information-theoretic and polynomial-time regret when tracking small expert subsets in online convex optimization. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Theoretical optimization paper, unrelated to deployment safety or AI guardrails.
Tracking a Small Set of Experts in Online Convex Optimization This paper proves a computational separation between information-theoretic and polynomial-time regret when tracking small expert subsets in online convex optimization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Theoretical online convex optimization regret bounds, not LLM evaluation.
Continual Learning for Reliable In-Sequence Molecular Communication This paper introduces a continual learning replay method to incrementally evaluate and adapt molecular communication models across changing experimental settings. Peer-reviewed	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Applies continual learning to bio-nanomachines, not small or efficient models.
Continual Learning for Reliable In-Sequence Molecular Communication This paper introduces a continual learning replay method to incrementally evaluate and adapt molecular communication models across changing experimental settings. Peer-reviewed	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on molecular networking simulations, not AI system deployment safety.
Continual Learning for Reliable In-Sequence Molecular Communication This paper introduces a continual learning replay method to incrementally evaluate and adapt molecular communication models across changing experimental settings. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Addresses molecular communication simulation, not language model evaluation.
The Accuracy Well: Nonmonotone Accuracy Dynamics at Vanishing Initialization This paper analyzes non-monotone accuracy dynamics during early neural network training when models are initialized with vanishingly small weights. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Explores optimization dynamics under vanishing initialization, not model efficiency or quantization.
The Accuracy Well: Nonmonotone Accuracy Dynamics at Vanishing Initialization This paper analyzes non-monotone accuracy dynamics during early neural network training when models are initialized with vanishingly small weights. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on general neural network training phenomena, not AI safety or deployment guardrails.
The Accuracy Well: Nonmonotone Accuracy Dynamics at Vanishing Initialization This paper analyzes non-monotone accuracy dynamics during early neural network training when models are initialized with vanishingly small weights. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Analyzes training dynamics in standard vision architectures, not LLM benchmarking.
What Does Disagreement Mean? A Semantic Identifiability Frontier for Pairwise Preference Learning This paper establishes theoretical limits on what forms of annotator disagreement can be identified from pairwise preference feedback data. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on preference learning theory, not small models or efficient inference.
Learning by Examples: Power Grid Steady State Analysis with a Tabular Foundation Model This paper evaluates whether a pretrained tabular foundation model can perform power grid steady-state analysis without task-specific retraining. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on power grid physical simulation accuracy, not AI deployment safety.
Learning by Examples: Power Grid Steady State Analysis with a Tabular Foundation Model This paper evaluates whether a pretrained tabular foundation model can perform power grid steady-state analysis without task-specific retraining. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Evaluates TabPFN on power grid analysis, not model compression or efficient serving.
Learning by Examples: Power Grid Steady State Analysis with a Tabular Foundation Model This paper evaluates whether a pretrained tabular foundation model can perform power grid steady-state analysis without task-specific retraining. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Studies tabular models for power grids, not LLM evaluation frameworks.
Decoupled Marginal Sharpening for Training-Free Inference-Time Scaling This paper proposes a decoupled marginal sharpening method to improve inference-time reasoning in language models without weight updates or reward models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Optimizes reasoning path probabilities rather than AI safety, guardrails, or red teaming.
Decoupled Marginal Sharpening for Training-Free Inference-Time Scaling This paper proposes a decoupled marginal sharpening method to improve inference-time reasoning in language models without weight updates or reward models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents an inference sampling technique, not an LLM benchmark or evaluation methodology.
Average-Reward Policy Identification from a Single Continuing Trajectory This paper provides matching upper and lower bounds for learning near-optimal policies from a single continuing trajectory in average-reward reinforcement learning. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on theoretical RL sample complexity, not LLM evaluation benchmarks.
Average-Reward Policy Identification from a Single Continuing Trajectory This paper provides matching upper and lower bounds for learning near-optimal policies from a single continuing trajectory in average-reward reinforcement learning. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Studies trajectory navigation complexity in theoretical RL, not model compression or serving.
Average-Reward Policy Identification from a Single Continuing Trajectory This paper provides matching upper and lower bounds for learning near-optimal policies from a single continuing trajectory in average-reward reinforcement learning. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses RL policy exploration bounds, not deployment safety, guardrails, or hallucination.
Graph-Enhanced Disease Trajectory Discovery and Temporal Endpoint Forecasting from Longitudinal Electronic Health Records This paper introduces a graph-enhanced framework to discover disease trajectories and forecast temporal clinical endpoints from electronic health records. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on clinical endpoint prediction, not LLM benchmarking or evaluation.
Graph-Enhanced Disease Trajectory Discovery and Temporal Endpoint Forecasting from Longitudinal Electronic Health Records This paper introduces a graph-enhanced framework to discover disease trajectories and forecast temporal clinical endpoints from electronic health records. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not study model compression, distillation, or efficient model serving.
Graph-Enhanced Disease Trajectory Discovery and Temporal Endpoint Forecasting from Longitudinal Electronic Health Records This paper introduces a graph-enhanced framework to discover disease trajectories and forecast temporal clinical endpoints from electronic health records. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on healthcare forecasting rather than AI safety or system guardrails.
A Two-Level Clinical Novelty Prioritization Framework for Disease Trajectories Discovered from Longitudinal Electronic Health Records A two-level prioritization framework evaluates semantic novelty and reliability of disease trajectory patterns extracted from electronic health records. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model compression, quantization, or efficient LLM inference.
A Two-Level Clinical Novelty Prioritization Framework for Disease Trajectories Discovered from Longitudinal Electronic Health Records A two-level prioritization framework evaluates semantic novelty and reliability of disease trajectory patterns extracted from electronic health records. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses clinical motif ranking rather than AI safety or LLM guardrails.
A Two-Level Clinical Novelty Prioritization Framework for Disease Trajectories Discovered from Longitudinal Electronic Health Records A two-level prioritization framework evaluates semantic novelty and reliability of disease trajectory patterns extracted from electronic health records. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on clinical EHR trajectory analysis, not LLM benchmarks or evaluation.
Bayesian Learning to Select Fair Classifiers The paper presents an online Thompson Sampling bandit policy to select classifiers that balance predictive accuracy and subgroup fairness under delayed feedback. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on online bandit selection, not model compression or serving efficiency.
The Physics Is the Network: Fast, Training-Free Neural Emulation PolyNets embeds continuous polynomial equations into recurrent neural structures to perform fast, training-free physics simulations with known parameters. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on scientific PDE solvers rather than LLM inference or distillation.
The Physics Is the Network: Fast, Training-Free Neural Emulation PolyNets embeds continuous polynomial equations into recurrent neural structures to perform fast, training-free physics simulations with known parameters. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on physics-based PDE emulation, not LLM benchmarking or evaluation.
The Physics Is the Network: Fast, Training-Free Neural Emulation PolyNets embeds continuous polynomial equations into recurrent neural structures to perform fast, training-free physics simulations with known parameters. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address deployment safety, red-teaming, or LLM guardrails.
Revisiting Simple Baselines for Relational Deep Learning The paper demonstrates that simple feature propagation and shallow classifiers yield competitive baselines compared to complex architectures in relational deep learning. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Evaluates relational graph baselines on RelBench, not LLM evaluation methods.
Revisiting Simple Baselines for Relational Deep Learning The paper demonstrates that simple feature propagation and shallow classifiers yield competitive baselines compared to complex architectures in relational deep learning. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Investigates relational graph architectures, not LLM compression or inference serving.
Revisiting Simple Baselines for Relational Deep Learning The paper demonstrates that simple feature propagation and shallow classifiers yield competitive baselines compared to complex architectures in relational deep learning. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not focus on AI safety, alignment, hallucination, or guardrails.
Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation DualGCN uses dual-branch graph knowledge distillation with topology-smoothed semantic priors to address item cold-start collapse in recommender systems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Does not study LLM evaluation, benchmark contamination, or LLM-as-a-judge.
Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation DualGCN uses dual-branch graph knowledge distillation with topology-smoothed semantic priors to address item cold-start collapse in recommender systems. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on graph collaborative filtering distillation, not LLM compression or efficiency.
Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation DualGCN uses dual-branch graph knowledge distillation with topology-smoothed semantic priors to address item cold-start collapse in recommender systems. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on recommender systems rather than retrieval-augmented generation for LLMs.
Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation DualGCN uses dual-branch graph knowledge distillation with topology-smoothed semantic priors to address item cold-start collapse in recommender systems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address AI safety, jailbreaking, red-teaming, or guardrails.
Creating SquashFS Images to Support HPC and AI Datasets The paper benchmarks eleven workflows for packing small-file datasets into SquashFS images to ease metadata loads on high-performance computing systems. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on HPC file system compression, not retrieval-augmented generation.
$λ_0$: Scaling Open-World Egocentric Human Data for Humanoid Loco-Manipulation HumanVerse provides an egocentric dataset and alignment pipeline to transfer human loco-manipulation behaviors directly to humanoid robots without target-task fine-tuning. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Addresses humanoid loco-manipulation robotics, not industrial visual quality control.
$λ_0$: Scaling Open-World Egocentric Human Data for Humanoid Loco-Manipulation HumanVerse provides an egocentric dataset and alignment pipeline to transfer human loco-manipulation behaviors directly to humanoid robots without target-task fine-tuning. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on robotic loco-manipulation policies rather than multimodal foundation models.
Action Jets: Learning the Local Shape of the Action-conditioned Data Manifold Action Jets models the local shape of action-conditioned manifolds via Taylor approximations to enable efficient short-horizon robotic planning. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on vision-action dynamics manifolds without multimodal language or document understanding.
Action Jets: Learning the Local Shape of the Action-conditioned Data Manifold Action Jets models the local shape of action-conditioned manifolds via Taylor approximations to enable efficient short-horizon robotic planning. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Investigates predictive world models for planning, not industrial inspection.
How Far Can Coding Agents Go in Visual Generation? The study evaluates coding agents generating images via executable code, analyzing their strengths in geometric constraints over photorealistic drawing. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model compression, quantization, distillation, or serving efficiency.

6,301–6,360 of 17,313← Previous Next →