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
MultiAlloc: Budget-Aware Precision Allocation for KV Cache Quantization The paper presents a profiling-guided framework to allocate precision budgets adaptively across layers and tokens during KV cache quantization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Proposes a KV cache quantization method, not an evaluation benchmark.
BridgeTwist: Twisting Schrödinger Bridges for Training-Free Conditional Sampling The paper introduces a training-free conditional sampling method for pretrained Schrödinger bridge generative models. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Addresses continuous transport sampling, not small or efficient language models.
BridgeTwist: Twisting Schrödinger Bridges for Training-Free Conditional Sampling The paper introduces a training-free conditional sampling method for pretrained Schrödinger bridge generative models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on conditional sampling algorithms, not LLM evaluation methods.
BridgeTwist: Twisting Schrödinger Bridges for Training-Free Conditional Sampling The paper introduces a training-free conditional sampling method for pretrained Schrödinger bridge generative models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on stochastic transport sampling, not AI deployment safety or guardrails.
A Refresh Interval Does Not Fully Specify a Refresh Schedule: Learning-Rate-Dependent Placement in Kronecker-Preconditioned Language Model Training The paper demonstrates that the placement timing of curvature updates in Kronecker-preconditioned optimizers significantly affects language model training loss. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on pretraining optimizer schedules, not deployment safety or robustness.
A Refresh Interval Does Not Fully Specify a Refresh Schedule: Learning-Rate-Dependent Placement in Kronecker-Preconditioned Language Model Training The paper demonstrates that the placement timing of curvature updates in Kronecker-preconditioned optimizers significantly affects language model training loss. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Studies optimizer refresh schedules, not LLM evaluation or benchmarking frameworks.
A Refresh Interval Does Not Fully Specify a Refresh Schedule: Learning-Rate-Dependent Placement in Kronecker-Preconditioned Language Model Training The paper demonstrates that the placement timing of curvature updates in Kronecker-preconditioned optimizers significantly affects language model training loss. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on optimizer scheduling during pretraining, not model compression or inference efficiency.
On Collusion-Free Teaching in Planar Graphs The paper proves theoretical teaching complexity bounds for closed neighborhoods in planar graphs without collusion. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Theoretical machine learning paper on graph teaching complexity, not LLM benchmarks.
On Collusion-Free Teaching in Planar Graphs The paper proves theoretical teaching complexity bounds for closed neighborhoods in planar graphs without collusion. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on theoretical teaching complexity, not model compression or efficiency.
On Collusion-Free Teaching in Planar Graphs The paper proves theoretical teaching complexity bounds for closed neighborhoods in planar graphs without collusion. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Theoretical learning theory paper, not AI safety or deployment robustness.
Learning Source Structure: Bayesian Prediction with Transformers The paper compares transformer sequence predictions against an optimal Bayesian predictor across structured Markov sources. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Studies theoretical sequence prediction, not model compression or efficient serving.
Learning Source Structure: Bayesian Prediction with Transformers The paper compares transformer sequence predictions against an optimal Bayesian predictor across structured Markov sources. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Theoretical study of synthetic sequence priors, not deployment safety or guardrails.
Learning Source Structure: Bayesian Prediction with Transformers The paper compares transformer sequence predictions against an optimal Bayesian predictor across structured Markov sources. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on theoretical Bayesian sequence prediction, not LLM evaluation benchmarks.
Learning When to Transfer: Reinforcement Learning for Carbon-Footprint Modeling in New Manufacturing Domains The paper proposes a reinforcement learning method to selectively transfer data for carbon-footprint modeling in new manufacturing settings. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Studies transfer learning for manufacturing metrics, not LLM evaluation.
Learning When to Transfer: Reinforcement Learning for Carbon-Footprint Modeling in New Manufacturing Domains The paper proposes a reinforcement learning method to selectively transfer data for carbon-footprint modeling in new manufacturing settings. Peer-reviewed	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Deals with domain transfer policies, not small or efficient LLM architectures.
Learning When to Transfer: Reinforcement Learning for Carbon-Footprint Modeling in New Manufacturing Domains The paper proposes a reinforcement learning method to selectively transfer data for carbon-footprint modeling in new manufacturing settings. Peer-reviewed	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on tabular carbon footprint prediction, not AI safety or LLM guardrails.
SP$^2$ec: Adaptive Self-Speculative Decoding for Vision-Language Models The paper presents an adaptive, training-free speculative decoding method to speed up inference for vision-language models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on inference acceleration techniques rather than model evaluation.
SP$^2$ec: Adaptive Self-Speculative Decoding for Vision-Language Models The paper presents an adaptive, training-free speculative decoding method to speed up inference for vision-language models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on decoding acceleration, not safety, jailbreaks, or robustness.
Enabling Token-wise Importance Sampling in Off-Policy RLVR with Proximal KL Regularization The paper develops an off-policy reinforcement learning method with token-wise importance sampling and proximal KL regularization. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on RL optimization algorithms rather than model compression or efficient serving.
Enabling Token-wise Importance Sampling in Off-Policy RLVR with Proximal KL Regularization The paper develops an off-policy reinforcement learning method with token-wise importance sampling and proximal KL regularization. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on reinforcement learning optimization algorithms rather than deployment safety or guardrails.
Enabling Token-wise Importance Sampling in Off-Policy RLVR with Proximal KL Regularization The paper develops an off-policy reinforcement learning method with token-wise importance sampling and proximal KL regularization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Proposes an RL training algorithm, not an evaluation methodology or benchmark.
Emergent Loop Geometry in Gauge-Equivariant Graph Neural Networks The paper analyzes how network depth influences the internal loop geometry and spatial reach of gauge-equivariant graph neural networks. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on theoretical physics GNN representations, not model efficiency.
Emergent Loop Geometry in Gauge-Equivariant Graph Neural Networks The paper analyzes how network depth influences the internal loop geometry and spatial reach of gauge-equivariant graph neural networks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on lattice gauge theory, not AI safety or reliability.
Emergent Loop Geometry in Gauge-Equivariant Graph Neural Networks The paper analyzes how network depth influences the internal loop geometry and spatial reach of gauge-equivariant graph neural networks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on mathematical properties of graph neural networks in physics.
Depth Memory The paper formalizes layer-to-layer representation sharing as a memory storage and retrieval framework for deep transformer networks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Proposes a model architecture framework, not an evaluation benchmark or method.
Depth Memory The paper formalizes layer-to-layer representation sharing as a memory storage and retrieval framework for deep transformer networks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on internal layer representation mixing, not safety, jailbreaks, or hallucinations.
Statistical Vignettes on the Grokking Phenomenon The paper analytically characterizes the mechanisms and conditions behind the grokking phenomenon in a simplified learning setup. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Theoretical analysis of grokking dynamics, not LLM evaluation or benchmarks.
Statistical Vignettes on the Grokking Phenomenon The paper analytically characterizes the mechanisms and conditions behind the grokking phenomenon in a simplified learning setup. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Theoretical learning study, not small models, pruning, or efficient inference.
Statistical Vignettes on the Grokking Phenomenon The paper analytically characterizes the mechanisms and conditions behind the grokking phenomenon in a simplified learning setup. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Studies theoretical generalization dynamics, not deployment safety or guardrails.
Joint Neural Posterior Estimation and Model Selection for Chemical Reaction Networks The paper presents a neural network architecture for simultaneous model selection and parameter posterior estimation in chemical reaction networks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on chemical parameter estimation, not AI safety or deployment reliability.
Joint Neural Posterior Estimation and Model Selection for Chemical Reaction Networks The paper presents a neural network architecture for simultaneous model selection and parameter posterior estimation in chemical reaction networks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on chemical reaction Bayesian inference, not LLM benchmarking.
Joint Neural Posterior Estimation and Model Selection for Chemical Reaction Networks The paper presents a neural network architecture for simultaneous model selection and parameter posterior estimation in chemical reaction networks. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Addresses Bayesian inference for chemistry, not small or efficient language models.
Exact Topological Compliance: Generating Persistence-Equivalent Graphs The paper introduces methods to generate graphs whose topological persistence diagrams match target descriptors exactly rather than approximately. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on topological graph generation, not model compression or efficient serving.
Exact Topological Compliance: Generating Persistence-Equivalent Graphs The paper introduces methods to generate graphs whose topological persistence diagrams match target descriptors exactly rather than approximately. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on topological graph theory, not deployment safety or reliability guardrails.
Exact Topological Compliance: Generating Persistence-Equivalent Graphs The paper introduces methods to generate graphs whose topological persistence diagrams match target descriptors exactly rather than approximately. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Addresses theoretical graph generation, not LLM benchmarking or evaluation methods.
Data Free Training of Diffusion Models for Lattice Field Theories The paper presents a method for training diffusion models in lattice field theory directly from their own generated samples. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on physics simulation methods, not AI safety, jailbreaks, or alignment.
Data Free Training of Diffusion Models for Lattice Field Theories The paper presents a method for training diffusion models in lattice field theory directly from their own generated samples. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Deals with physics sampling in lattice field theory, not LLM evaluation.
Data Free Training of Diffusion Models for Lattice Field Theories The paper presents a method for training diffusion models in lattice field theory directly from their own generated samples. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Studies lattice field theory sampling rather than efficient LLM inference or distillation.
Understanding and Mitigating Training Instability in Large-Scale Recommender Systems: A Cascade Perspective The paper analyzes training instability cascades in large recommender systems and proposes targeted gradient, weight, and loss interventions. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Analyzes training convergence instability rather than LLM benchmarking or evaluation frameworks.
Understanding and Mitigating Training Instability in Large-Scale Recommender Systems: A Cascade Perspective The paper analyzes training instability cascades in large recommender systems and proposes targeted gradient, weight, and loss interventions. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Addresses recommendation system training instability, not retrieval-augmented generation systems.
Understanding and Mitigating Training Instability in Large-Scale Recommender Systems: A Cascade Perspective The paper analyzes training instability cascades in large recommender systems and proposes targeted gradient, weight, and loss interventions. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on recommender training divergence rather than model compression or efficient inference.
EVISTA: Unified Multimodal Video Search with Evidence-Aware Temporal Path Optimization The paper introduces a multimodal video retrieval system that optimizes temporal paths to locate ordered events across video, text, and audio. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Focuses on multimodal video keyframe search, not RAG or knowledge-intensive LLM retrieval.
A Stateful Research Information Infrastructure for Scientific Discovery The paper introduces an infrastructure that organizes scientific works as persistent, stateful research objects to maintain continuity across tools. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Presents data infrastructure for scientific objects, not retrieval-augmented generation or grounding mechanisms.
MoSAIC: Learning DAG Scheduling Heuristics from Optimal Schedules through Motif-Aware Recovery and LLM-Guided Reasoning The paper introduces a framework that uses language models to extract task-scheduling priority functions from optimal solver solutions on directed acyclic graphs. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on task scheduling, not deployment safety, jailbreaks, or hallucinations.
MoSAIC: Learning DAG Scheduling Heuristics from Optimal Schedules through Motif-Aware Recovery and LLM-Guided Reasoning The paper introduces a framework that uses language models to extract task-scheduling priority functions from optimal solver solutions on directed acyclic graphs. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on DAG scheduling optimization, not LLM evaluation or benchmark methodology.
MoSAIC: Learning DAG Scheduling Heuristics from Optimal Schedules through Motif-Aware Recovery and LLM-Guided Reasoning The paper introduces a framework that uses language models to extract task-scheduling priority functions from optimal solver solutions on directed acyclic graphs. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Deals with DAG task scheduling, not model compression or efficient serving.
MoSAIC: Learning DAG Scheduling Heuristics from Optimal Schedules through Motif-Aware Recovery and LLM-Guided Reasoning The paper introduces a framework that uses language models to extract task-scheduling priority functions from optimal solver solutions on directed acyclic graphs. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Focuses on DAG scheduling heuristics rather than text retrieval or knowledge grounding.
Toward Explainability in Multimodal Recommendation The paper reviews explainability methods and evaluation metrics in multimodal recommender systems, categorizing approaches into modality and post-hoc explanations. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Surveys explainability in multimodal recommender systems, not retrieval-augmented generation.
Federated Semantic Residuals for Multimodal Recommendation The paper proposes a federated recommendation adapter that adds bounded semantic residual scores to on-device collaborative filtering models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Studies federated multimodal recommender systems rather than retrieval-augmented generation.
Pathwise Alignment for Collaborative Filtering: Bilateral Conditional Flow Regularization The paper uses conditional flow matching during training to regularize matrix factorization embeddings across sparse user interactions in collaborative filtering. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Deals with collaborative filtering algorithms, not retrieval-augmented generation systems.
Rethinking Evidence Modeling: Context-Coupled Graph Learning for Multimodal Recommendation The paper presents a context-coupled graph learning method that jointly models collaborative and multimodal item relations for recommendation systems. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Focuses on graph-based multimodal recommendation systems rather than retrieval-augmented generation.
UniSGR-V2: Stepwise Target-Side Feature Anticipation and Crossing for Generative Recommendation The paper proposes a generative recommendation framework that reconstructs continuous target-side features during autoregressive decoding to improve ranking accuracy. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Focuses on generative recommendation systems, not retrieval-augmented generation for language models.
Tackling the Modality Inhibitance Effect with Wavelet-based Modality Integration in Multi-modal Recommendation The paper introduces a wavelet-based integration method to resolve performance degradation caused by feature redundancy in trimodal recommender systems. Peer-reviewed	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Focuses on multimodal recommendation algorithms rather than retrieval-augmented generation for language models.
Tackling the Modality Inhibitance Effect with Wavelet-based Modality Integration in Multi-modal Recommendation The paper introduces a wavelet-based integration method to resolve performance degradation caused by feature redundancy in trimodal recommender systems. Peer-reviewed	Multimodal models via arxiv · 15 Oct, 06:18	Not relevant Deals with recommender feature fusion, not general multimodal models like vision-language foundation models.
From a Large-Scale Teacher to Efficient Multimodal Students for Search Relevance Prediction The paper distills an LLM-based search relevance judge into smaller multimodal student models for efficient e-commerce search evaluation. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on model distillation for relevance prediction, not AI safety, jailbreaks, or guardrails.
From a Large-Scale Teacher to Efficient Multimodal Students for Search Relevance Prediction The paper distills an LLM-based search relevance judge into smaller multimodal student models for efficient e-commerce search evaluation. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Studies e-commerce search relevance prediction rather than retrieval-augmented generation.
Tracing Money Flows in the Multi-Token Cryptocurrency Network The paper develops a transaction-aware graph method to trace multi-token fund flows and asset transformations across cryptocurrency networks. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Analyzes cryptocurrency transaction network graphs, not retrieval-augmented generation or document search.
The Scored Population, the Operating Point, and the Pipeline: Three Objects That Decide What an Open-Web Event System Looks Like The paper studies how pipeline selection and scored populations introduce evaluation blind spots in open-web event admission systems. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Analyzes web event extraction and admission pipelines rather than grounding language models with external knowledge.
Mitigating Popularity Bias in Sequential Recommendations with Sparse Prior-Regularized Objective The paper proposes a sparse prior-regularized objective that makes popularity bias correction context-dependent during sequential recommender model training. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18	Not relevant Focuses on recommender systems popularity bias rather than retrieval-augmented generation for language models.
Human Pose Estimation by Probabilistic Mixtures of Gaussian and Uniform Distributions The paper proposes a continuous probabilistic Gaussian mixture framework to improve human pose estimation under dense occlusions without retraining models. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on probabilistic pose estimation modeling rather than model compression or efficient serving.

9,541–9,600 of 17,313← Previous Next →