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
Radial Tail Repair: Post-Training Correction for Diffusion and Flow Models The paper proposes a post-training tail correction method for diffusion and flow models using radial-angular extreme value decomposition. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on distribution tail modeling, not model compression.
GpFlash: Maximizing Speed and Minimizing Memory for Gaussian Process Training and Prediction The paper introduces GpFlash, an open library designed to speed up Gaussian process training and prediction while reducing memory footprint. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Deals with Gaussian process computing, not AI deployment safety.
GpFlash: Maximizing Speed and Minimizing Memory for Gaussian Process Training and Prediction The paper introduces GpFlash, an open library designed to speed up Gaussian process training and prediction while reducing memory footprint. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on Gaussian processes rather than small language models.
GpFlash: Maximizing Speed and Minimizing Memory for Gaussian Process Training and Prediction The paper introduces GpFlash, an open library designed to speed up Gaussian process training and prediction while reducing memory footprint. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Covers Gaussian process libraries, not LLM benchmark evaluation.
Better together? A performance theory of human-machine skill policies under generalization difficulty The paper develops a formal performance theory analyzing skill allocation and workflow design in joint human-machine systems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on theoretical workflow economics rather than deployment safety.
Better together? A performance theory of human-machine skill policies under generalization difficulty The paper develops a formal performance theory analyzing skill allocation and workflow design in joint human-machine systems. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Covers task allocation theory, not model compression or efficiency.
Better together? A performance theory of human-machine skill policies under generalization difficulty The paper develops a formal performance theory analyzing skill allocation and workflow design in joint human-machine systems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Presents theoretical framework for human-machine pairing, not LLM benchmarks.
Efficient Resource-Shiftable RL Post-training with Generative Rewards The paper presents PivotRL, a system that dynamically shifts GPU resources among rollout, reward modeling, and training during post-training. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on cluster systems engineering rather than AI safety.
Efficient Resource-Shiftable RL Post-training with Generative Rewards The paper presents PivotRL, a system that dynamically shifts GPU resources among rollout, reward modeling, and training during post-training. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Addresses training cluster resource scheduling rather than benchmark evaluation.
From Parallel to Sequential: Small-Batch Noise Restores Simplicity Bias of Muon The paper demonstrates that small-batch stochastic gradient noise restores sequential feature learning in the Muon optimizer. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Analyzes optimizer dynamics rather than language model evaluation methods.
From Parallel to Sequential: Small-Batch Noise Restores Simplicity Bias of Muon The paper demonstrates that small-batch stochastic gradient noise restores sequential feature learning in the Muon optimizer. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Studies optimizer simplicity bias rather than deployment safety.
From Parallel to Sequential: Small-Batch Noise Restores Simplicity Bias of Muon The paper demonstrates that small-batch stochastic gradient noise restores sequential feature learning in the Muon optimizer. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on optimizer theory rather than inference efficiency.
Sequence Composition in EHR Foundation Model Pretraining The paper analyzes sequence composition strategies during electronic health record foundation model pretraining to improve downstream clinical prediction. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Explores EHR pretraining sequences rather than general LLM benchmarking.
Sequence Composition in EHR Foundation Model Pretraining The paper analyzes sequence composition strategies during electronic health record foundation model pretraining to improve downstream clinical prediction. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Studies pretraining data composition rather than model compression methods.
Sequence Composition in EHR Foundation Model Pretraining The paper analyzes sequence composition strategies during electronic health record foundation model pretraining to improve downstream clinical prediction. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Addresses EHR pretraining sequences rather than AI safety guardrails.
Learning Operators for Symmetry Discovery: From Lie Groups to Local Symmetry Geometry The paper formulates continuous symmetry discovery across function families as an inverse symmetry operator problem using differential characteristics. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Studies mathematical continuous symmetry discovery, not LLM evaluation.
Learning Operators for Symmetry Discovery: From Lie Groups to Local Symmetry Geometry The paper formulates continuous symmetry discovery across function families as an inverse symmetry operator problem using differential characteristics. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on operator theory rather than deployment safety.
Learning Operators for Symmetry Discovery: From Lie Groups to Local Symmetry Geometry The paper formulates continuous symmetry discovery across function families as an inverse symmetry operator problem using differential characteristics. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Deals with Lie groups and operators, not model efficiency.
FORGE: Procedural Weight Generation for Sparse MoE Inference The paper introduces FORGE, an on-chip datapath that regenerates sparse mixture-of-experts weights from compact seeds during inference. Peer-reviewed	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on hardware inference acceleration rather than safety risks.
FORGE: Procedural Weight Generation for Sparse MoE Inference The paper introduces FORGE, an on-chip datapath that regenerates sparse mixture-of-experts weights from compact seeds during inference. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Proposes hardware weight generation rather than LLM benchmarking methods.
AI-Assisted Tumor-Growth Law Discovery from Sparse Data: Refitting and LLM-Based Ranking The paper evaluates symbolic regression, refitting techniques, and language model rankings for inferring tumor-growth laws from sparse observation data. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Studies symbolic regression fitting rather than AI safety guardrails.
AI-Assisted Tumor-Growth Law Discovery from Sparse Data: Refitting and LLM-Based Ranking The paper evaluates symbolic regression, refitting techniques, and language model rankings for inferring tumor-growth laws from sparse observation data. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on tumor growth discovery rather than model compression.
ML Engineering Agents - A Data Management Perspective The tutorial examines machine learning engineering agents from a data management view, highlighting challenges in pipeline execution, data discovery, and debugging. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Discusses pipeline automation agents, not model compression or serving.
ML Engineering Agents - A Data Management Perspective The tutorial examines machine learning engineering agents from a data management view, highlighting challenges in pipeline execution, data discovery, and debugging. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Covers ML pipeline agents rather than LLM benchmark design.
ML Engineering Agents - A Data Management Perspective The tutorial examines machine learning engineering agents from a data management view, highlighting challenges in pipeline execution, data discovery, and debugging. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Covers data engineering for ML agents, not core AI safety.
Time Window Optimization for Motor Imagery Classification via Temporal Structure Embedded Reinforcement Learning The paper presents a temporal reinforcement learning framework that selects physiologically aligned time windows for motor imagery brain-computer interfaces. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Addresses BCI classification rather than AI safety or reliability.
Time Window Optimization for Motor Imagery Classification via Temporal Structure Embedded Reinforcement Learning The paper presents a temporal reinforcement learning framework that selects physiologically aligned time windows for motor imagery brain-computer interfaces. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on EEG signal processing, not LLM benchmark evaluation.
Time Window Optimization for Motor Imagery Classification via Temporal Structure Embedded Reinforcement Learning The paper presents a temporal reinforcement learning framework that selects physiologically aligned time windows for motor imagery brain-computer interfaces. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on brain-computer interfaces, not model compression or efficiency.
Minimax Optimality of Shampoo for Nonstationary Stochastic Optimization The paper provides theoretical convergence and tracking error bounds for Matrix Shampoo optimization in nonstationary stochastic settings. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Theoretical optimizer bounds, not AI safety and reliability.
Minimax Optimality of Shampoo for Nonstationary Stochastic Optimization The paper provides theoretical convergence and tracking error bounds for Matrix Shampoo optimization in nonstationary stochastic settings. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Mathematical optimization theory, not small or efficient models.
Minimax Optimality of Shampoo for Nonstationary Stochastic Optimization The paper provides theoretical convergence and tracking error bounds for Matrix Shampoo optimization in nonstationary stochastic settings. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Theoretical convergence of an optimization algorithm, not LLM evaluation.
Making Tensor Decoupling More Robust and Less Parametric The paper provides two modifications to make tensor decoupling methods more robust to magnitude mismatches and eliminate manual spline tuning. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Addresses spline-based tensor decoupling, not model compression.
Making Tensor Decoupling More Robust and Less Parametric The paper provides two modifications to make tensor decoupling methods more robust to magnitude mismatches and eliminate manual spline tuning. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on mathematical function approximation, not deployment safety.
Making Tensor Decoupling More Robust and Less Parametric The paper provides two modifications to make tensor decoupling methods more robust to magnitude mismatches and eliminate manual spline tuning. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Covers tensor decoupling algorithms, not LLM evaluation.
Learning to Design Convex Relaxations for Nonconvex Optimization The paper presents AutoConvexRelax, which learns a reinforcement learning policy to automatically construct valid convex relaxations for nonconvex optimization problems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Studies mathematical relaxation proofs, not AI deployment safety.
Learning to Design Convex Relaxations for Nonconvex Optimization The paper presents AutoConvexRelax, which learns a reinforcement learning policy to automatically construct valid convex relaxations for nonconvex optimization problems. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on mathematical optimization surrogates, not small models.
Learning to Design Convex Relaxations for Nonconvex Optimization The paper presents AutoConvexRelax, which learns a reinforcement learning policy to automatically construct valid convex relaxations for nonconvex optimization problems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Deals with convex relaxations in optimization, not LLM evaluation.
DiTSG: Context-Aware Air Traffic Scene Generation with Multi-Agent Diffusion Transformers The paper proposes a multi-agent diffusion transformer framework to generate realistic air traffic scenes and trajectory patterns from recorded data. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Diffusion transformer for trajectory data, not model efficiency.
DiTSG: Context-Aware Air Traffic Scene Generation with Multi-Agent Diffusion Transformers The paper proposes a multi-agent diffusion transformer framework to generate realistic air traffic scenes and trajectory patterns from recorded data. Preprint	AI agents via arxiv · 15 Oct, 06:18	Not relevant Focuses on trajectory generation models, not multi-step LLM agents.
DiTSG: Context-Aware Air Traffic Scene Generation with Multi-Agent Diffusion Transformers The paper proposes a multi-agent diffusion transformer framework to generate realistic air traffic scenes and trajectory patterns from recorded data. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Generates traffic scenes, not LLM benchmarking or evaluation.
DiTSG: Context-Aware Air Traffic Scene Generation with Multi-Agent Diffusion Transformers The paper proposes a multi-agent diffusion transformer framework to generate realistic air traffic scenes and trajectory patterns from recorded data. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Studies air traffic trajectories, not industrial computer vision.
DiTSG: Context-Aware Air Traffic Scene Generation with Multi-Agent Diffusion Transformers The paper proposes a multi-agent diffusion transformer framework to generate realistic air traffic scenes and trajectory patterns from recorded data. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Simulates airspace scenarios, not AI deployment safety or guardrails.
Learning to Design SR Solvers: Harness Optimization for Symbolic Regression The paper introduces HarnessSR to jointly optimize agent workflows and prompts for discovering symbolic regression equations from data. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on optimizing symbolic regression agents, not LLM benchmarks.
Learning to Design SR Solvers: Harness Optimization for Symbolic Regression The paper introduces HarnessSR to jointly optimize agent workflows and prompts for discovering symbolic regression equations from data. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Studies prompt optimization for regression, not deployment safety.
Learning to Design SR Solvers: Harness Optimization for Symbolic Regression The paper introduces HarnessSR to jointly optimize agent workflows and prompts for discovering symbolic regression equations from data. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Addresses symbolic regression workflows, not efficient or small models.
Generative Models as Adaptive Trust Regions for Discrete Bayesian Optimisation The paper presents GenTuRBO, using pretrained generative models as adaptive proposal trust regions for discrete Bayesian optimization tasks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Studies discrete Bayesian optimisation, not LLM benchmark evaluation.
Generative Models as Adaptive Trust Regions for Discrete Bayesian Optimisation The paper presents GenTuRBO, using pretrained generative models as adaptive proposal trust regions for discrete Bayesian optimization tasks. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Optimization framework for discrete designs, not model efficiency.
Generative Models as Adaptive Trust Regions for Discrete Bayesian Optimisation The paper presents GenTuRBO, using pretrained generative models as adaptive proposal trust regions for discrete Bayesian optimization tasks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on discrete Bayesian optimization, not deployment safety.
Constant-Depth Rational-Weight GNNs and their Connection to Schanuel's Conjecture The paper proves constant-depth rational-weight graph neural networks match the 1-WL expressiveness bound under Schanuel's conjecture from number theory. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on theoretical GNN expressiveness, not model compression.
Constant-Depth Rational-Weight GNNs and their Connection to Schanuel's Conjecture The paper proves constant-depth rational-weight graph neural networks match the 1-WL expressiveness bound under Schanuel's conjecture from number theory. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Theoretical graph expressiveness proof, not deployment safety.
Constant-Depth Rational-Weight GNNs and their Connection to Schanuel's Conjecture The paper proves constant-depth rational-weight graph neural networks match the 1-WL expressiveness bound under Schanuel's conjecture from number theory. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Theoretical study of GNN expressiveness, not LLM evaluation.
Intrinsic Geometry Matching: learning observation-equivariant process geometry The paper proposes a method to learn the intrinsic geometry and diffusion operators of underlying processes from distorted, indirect observations. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on geometric representation learning, not AI safety.
Intrinsic Geometry Matching: learning observation-equivariant process geometry The paper proposes a method to learn the intrinsic geometry and diffusion operators of underlying processes from distorted, indirect observations. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Theoretical manifold learning method, not model efficiency.
Intrinsic Geometry Matching: learning observation-equivariant process geometry The paper proposes a method to learn the intrinsic geometry and diffusion operators of underlying processes from distorted, indirect observations. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Presents differential geometry learning, not LLM evaluation.
A physics-coupled graph neural operator for mid-plane six-component stress tensor field prediction in concrete penetration The paper introduces a physics-coupled graph neural operator to predict mid-plane stress tensor fields in concrete during projectile penetration. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Structural engineering physics surrogate, not AI deployment safety.
A physics-coupled graph neural operator for mid-plane six-component stress tensor field prediction in concrete penetration The paper introduces a physics-coupled graph neural operator to predict mid-plane stress tensor fields in concrete during projectile penetration. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Predicts concrete stress fields, not LLM benchmark evaluation.
A physics-coupled graph neural operator for mid-plane six-component stress tensor field prediction in concrete penetration The paper introduces a physics-coupled graph neural operator to predict mid-plane stress tensor fields in concrete during projectile penetration. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Neural operator for materials physics, not efficient models.
Multi-Agent Deferred-Decision Trajectory Optimization with Anytime-Valid Target Retention for Moving Candidates This paper develops a deferred-decision trajectory optimization controller with conformal prediction guarantees for multi-agent target interception. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Addresses robotics trajectory optimization, not LLM evaluation.
Multi-Agent Deferred-Decision Trajectory Optimization with Anytime-Valid Target Retention for Moving Candidates This paper develops a deferred-decision trajectory optimization controller with conformal prediction guarantees for multi-agent target interception. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Studies target interception control, not AI safety and reliability.
Multi-Agent Deferred-Decision Trajectory Optimization with Anytime-Valid Target Retention for Moving Candidates This paper develops a deferred-decision trajectory optimization controller with conformal prediction guarantees for multi-agent target interception. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:18	Not relevant Focuses on trajectory interception planning, not industrial visual inspection.

9,121–9,180 of 17,313← Previous Next →