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
Beyond Task Success: A Theory of Representation Reuse This paper develops a theoretical framework characterizing when representations learned on a source task remain recoverable and reusable for future tasks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Presents general representation learning theory, not LLM evaluation benchmarks.
Beyond RMSE and Granger Gain: Multi-Horizon Forecast-Error Entropy for Nonlinear Predictive Coupling This paper proposes a forecast-error distribution entropy framework to detect nonlinear predictive coupling across multiple time horizons. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Covers time-series entropy bounds, not deployment safety or reliability.
Beyond RMSE and Granger Gain: Multi-Horizon Forecast-Error Entropy for Nonlinear Predictive Coupling This paper proposes a forecast-error distribution entropy framework to detect nonlinear predictive coupling across multiple time horizons. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on time-series forecasting metrics, not model efficiency.
Beyond RMSE and Granger Gain: Multi-Horizon Forecast-Error Entropy for Nonlinear Predictive Coupling This paper proposes a forecast-error distribution entropy framework to detect nonlinear predictive coupling across multiple time horizons. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Studies time-series predictive causality metrics, not language model evaluation.
LunaTP: Enabling Efficient and Scalable Tensor Product for Equivariant MLIPs This paper introduces LunaTP, a parallel execution framework that accelerates tensor product convolutions in equivariant machine learning interatomic potentials. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Hardware acceleration of tensor products, not LLM evaluation.
LunaTP: Enabling Efficient and Scalable Tensor Product for Equivariant MLIPs This paper introduces LunaTP, a parallel execution framework that accelerates tensor product convolutions in equivariant machine learning interatomic potentials. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Scientific computing acceleration, not AI deployment safety.
LunaTP: Enabling Efficient and Scalable Tensor Product for Equivariant MLIPs This paper introduces LunaTP, a parallel execution framework that accelerates tensor product convolutions in equivariant machine learning interatomic potentials. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on equivariant interatomic simulations, not small/efficient LLMs.
Overcoming Catastrophic Forgetting: Continual Learning for Radar Waveform Classification This paper benchmarks eighteen continual learning methods on a sequential radar waveform recognition task across task- and class-incremental settings. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Continual learning on radar datasets, not LLM benchmarks.
Overcoming Catastrophic Forgetting: Continual Learning for Radar Waveform Classification This paper benchmarks eighteen continual learning methods on a sequential radar waveform recognition task across task- and class-incremental settings. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Continual learning for radar waveforms, not efficient LLM serving.
Overcoming Catastrophic Forgetting: Continual Learning for Radar Waveform Classification This paper benchmarks eighteen continual learning methods on a sequential radar waveform recognition task across task- and class-incremental settings. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Domain-specific radar signal classification, not AI deployment safety.
Convergence-Rate-Guided Algorithm Design of Expected Improvement for Continuous Simulation Optimization This paper develops finite-sample bounds to guide algorithm design for Expected Improvement in continuous simulation optimization with heteroscedastic noise. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Stochastic kriging optimization algorithms, not model efficiency.
Convergence-Rate-Guided Algorithm Design of Expected Improvement for Continuous Simulation Optimization This paper develops finite-sample bounds to guide algorithm design for Expected Improvement in continuous simulation optimization with heteroscedastic noise. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Simulation optimization bounds, not AI deployment safety.
Convergence-Rate-Guided Algorithm Design of Expected Improvement for Continuous Simulation Optimization This paper develops finite-sample bounds to guide algorithm design for Expected Improvement in continuous simulation optimization with heteroscedastic noise. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Continuous simulation optimization theory, not LLM evaluation.
FastEEP: Fast Intra-Instance Elastic Scaling for Expert-Parallel MoE Serving This paper presents FastEEP, a system that speeds up intra-instance elastic scaling for Mixture-of-Experts serving by reusing runtime state. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Serving infrastructure system design, not benchmark or model evaluation.
FastEEP: Fast Intra-Instance Elastic Scaling for Expert-Parallel MoE Serving This paper presents FastEEP, a system that speeds up intra-instance elastic scaling for Mixture-of-Experts serving by reusing runtime state. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Covers inference serving systems, not safety, guardrails, or alignment.
Graph Metanetworks for Transformer Based Language Models This paper develops graph metanetworks that respect the weight-space symmetries of transformer language models, including attention matrices and token embeddings. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on weight-space metanetworks, not LLM evaluation benchmarks.
Graph Metanetworks for Transformer Based Language Models This paper develops graph metanetworks that respect the weight-space symmetries of transformer language models, including attention matrices and token embeddings. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Metanetworks on model weights, not AI deployment safety.
Graph Metanetworks for Transformer Based Language Models This paper develops graph metanetworks that respect the weight-space symmetries of transformer language models, including attention matrices and token embeddings. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Weight-space symmetry modeling, not model compression or efficiency.
ProTok: Unified Protein Tokenization via Modular Multi-Codebook Quantization This paper introduces ProTok, a unified discrete tokenizer for protein sequence, structure, and functional annotations using modular multi-codebook quantization. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Biological tokenization, not AI safety or deployment robustness.
ProTok: Unified Protein Tokenization via Modular Multi-Codebook Quantization This paper introduces ProTok, a unified discrete tokenizer for protein sequence, structure, and functional annotations using modular multi-codebook quantization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Evaluates protein tokenizers, not LLM evaluation methodologies.
ProTok: Unified Protein Tokenization via Modular Multi-Codebook Quantization This paper introduces ProTok, a unified discrete tokenizer for protein sequence, structure, and functional annotations using modular multi-codebook quantization. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Protein representation tokenization, not small or efficient models.
Lifting: Quantum State Discrimination Framework for Mid-Circuit Measurements This paper presents Lifting, a framework combining repetition encoding and neural decoding for mid-circuit quantum state discrimination on superconducting hardware. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Quantum measurement framework, not efficient model architectures.
Lifting: Quantum State Discrimination Framework for Mid-Circuit Measurements This paper presents Lifting, a framework combining repetition encoding and neural decoding for mid-circuit quantum state discrimination on superconducting hardware. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Hardware-level quantum state discrimination, not LLM evaluation.
Lifting: Quantum State Discrimination Framework for Mid-Circuit Measurements This paper presents Lifting, a framework combining repetition encoding and neural decoding for mid-circuit quantum state discrimination on superconducting hardware. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Superconducting quantum hardware reliability, not AI deployment safety.
SoK: Rethinking Membership Inference in Tabular Machine Learning This paper provides a systematization of knowledge on membership inference attacks in tabular machine learning across small-to-medium datasets. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Evaluates membership inference attacks, not LLM capabilities or benchmarks.
SoK: Rethinking Membership Inference in Tabular Machine Learning This paper provides a systematization of knowledge on membership inference attacks in tabular machine learning across small-to-medium datasets. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Privacy evaluation of tabular models, not model efficiency.
Recalibrating Many Questions with Few Labels: Minimax Rates and Identification Limits This paper provides minimax rates and limits for recalibrating predictions across many binary questions when each question has few ground-truth labels. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Statistical calibration analysis, not efficient serving or compression.
Recalibrating Many Questions with Few Labels: Minimax Rates and Identification Limits This paper provides minimax rates and limits for recalibrating predictions across many binary questions when each question has few ground-truth labels. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Theoretical probability recalibration bounds, not LLM benchmark testing.
Recalibrating Many Questions with Few Labels: Minimax Rates and Identification Limits This paper provides minimax rates and limits for recalibrating predictions across many binary questions when each question has few ground-truth labels. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Statistical recalibration rates, not AI safety or jailbreak defense.
Steering the Fidelity in State Preparation on Noisy Bosonic Systems with Risk-Seeking Optimization This paper introduces a risk-seeking optimization method to prepare target quantum states with high fidelity on noisy bosonic quantum hardware. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Quantum physics state preparation, unrelated to AI deployment safety.
Steering the Fidelity in State Preparation on Noisy Bosonic Systems with Risk-Seeking Optimization This paper introduces a risk-seeking optimization method to prepare target quantum states with high fidelity on noisy bosonic quantum hardware. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Deals with quantum state optimization, not LLM evaluation.
Steering the Fidelity in State Preparation on Noisy Bosonic Systems with Risk-Seeking Optimization This paper introduces a risk-seeking optimization method to prepare target quantum states with high fidelity on noisy bosonic quantum hardware. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Focuses on quantum computing states, not model efficiency.
How Low Can You Go: Fine-Tuning LLMs with Bayesian Optimization (LlMBO) This paper proposes LlMBO, applying Bayesian optimization to ultra-low-dimensional fine-tuning as an alternative to gradient-based policy optimization. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on parameter-efficient tuning, not safety or guardrails.
How Low Can You Go: Fine-Tuning LLMs with Bayesian Optimization (LlMBO) This paper proposes LlMBO, applying Bayesian optimization to ultra-low-dimensional fine-tuning as an alternative to gradient-based policy optimization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Presents a fine-tuning optimization method, not benchmark design.
Approximating Optimal Policies in Heterogeneous Asynchronous Multi-player Bandits This paper analyzes heterogeneous asynchronous multi-player bandits, proves the centralized allocation problem is NP-hard, and studies greedy approximation policies. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Multi-armed bandit optimization theory, not LLM evaluation.
Approximating Optimal Policies in Heterogeneous Asynchronous Multi-player Bandits This paper analyzes heterogeneous asynchronous multi-player bandits, proves the centralized allocation problem is NP-hard, and studies greedy approximation policies. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Theoretical bandit resource allocation, not AI safety guardrails.
Approximating Optimal Policies in Heterogeneous Asynchronous Multi-player Bandits This paper analyzes heterogeneous asynchronous multi-player bandits, proves the centralized allocation problem is NP-hard, and studies greedy approximation policies. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Multi-player bandit policies, not small or efficient LLMs.
Two For The Price Of One: Curvature-Aware Pruning Preserves Influence Functions Too This paper introduces CACTUS, an ADMM-based optimizer that speeds up curvature-aware pruning and preserves influence functions in compressed language models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on pruning techniques, not benchmark evaluation methodology.
Two For The Price Of One: Curvature-Aware Pruning Preserves Influence Functions Too This paper introduces CACTUS, an ADMM-based optimizer that speeds up curvature-aware pruning and preserves influence functions in compressed language models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on model compression speed and perplexity, not safety.
Euclidean Distance-Based Flow Matching This paper replaces squared Euclidean distance with Euclidean distance in Flow Matching to improve distributional robustness against data contamination. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Theoretical robustness against distribution contamination, not AI deployment safety.
Euclidean Distance-Based Flow Matching This paper replaces squared Euclidean distance with Euclidean distance in Flow Matching to improve distributional robustness against data contamination. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Flow matching theoretical objective, not LLM benchmark evaluation.
Euclidean Distance-Based Flow Matching This paper replaces squared Euclidean distance with Euclidean distance in Flow Matching to improve distributional robustness against data contamination. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Generative flow matching method, not model efficiency techniques.
Understanding Circulant Permutation to Extend the CMinHash Estimator The paper derives explicit variance approximations to extend circulant permutation techniques to different MinHash estimators. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Algorithmic hashing theory, not small or efficient LLMs.
Understanding Circulant Permutation to Extend the CMinHash Estimator The paper derives explicit variance approximations to extend circulant permutation techniques to different MinHash estimators. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on MinHash variance properties, not AI safety.
Understanding Circulant Permutation to Extend the CMinHash Estimator The paper derives explicit variance approximations to extend circulant permutation techniques to different MinHash estimators. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Variance analysis for MinHash, not LLM evaluation.
Last-Iterate Convergence of Clipped SGD with Linear Decay The paper provides last-iterate convergence bounds for clipped stochastic gradient descent under heavy-tailed noise and causal adversarial perturbations. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on theoretical SGD convergence under noise, not safety.
Last-Iterate Convergence of Clipped SGD with Linear Decay The paper provides last-iterate convergence bounds for clipped stochastic gradient descent under heavy-tailed noise and causal adversarial perturbations. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Convergence bounds for optimization, not model compression or efficiency.
Last-Iterate Convergence of Clipped SGD with Linear Decay The paper provides last-iterate convergence bounds for clipped stochastic gradient descent under heavy-tailed noise and causal adversarial perturbations. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Optimization theory for SGD, not LLM benchmark evaluation.
Linear Last-Iterate Convergence Rates for Optimistic Learning in Harmonic Games The paper establishes linear last-iterate convergence rates for optimistic learning algorithms in multiplayer harmonic games. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Game-theoretic convergence proof, unrelated to LLM evaluation.
Linear Last-Iterate Convergence Rates for Optimistic Learning in Harmonic Games The paper establishes linear last-iterate convergence rates for optimistic learning algorithms in multiplayer harmonic games. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Theoretical game theory analysis, not efficient model deployment.
Linear Last-Iterate Convergence Rates for Optimistic Learning in Harmonic Games The paper establishes linear last-iterate convergence rates for optimistic learning algorithms in multiplayer harmonic games. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Deals with game-theoretic equilibria, not AI deployment safety.
Fairness under Prediction Error in Predict-then-Optimize Pipelines The paper analyzes how prediction errors affect demographic parity fairness violations in budget-constrained allocation problems. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Addresses algorithmic fairness, not model compression or efficiency.
Fairness under Prediction Error in Predict-then-Optimize Pipelines The paper analyzes how prediction errors affect demographic parity fairness violations in budget-constrained allocation problems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Studies optimization fairness pipelines, not language model evaluation.
Deterministic Prediction as Global Velocity: Flow Matching for Probabilistic Time-Series Forecast The paper proposes a flow-matching model that integrates deterministic forecasting as a velocity component for stable probabilistic time-series predictions. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on time-series generative modeling, not LLM evaluation.
Deterministic Prediction as Global Velocity: Flow Matching for Probabilistic Time-Series Forecast The paper proposes a flow-matching model that integrates deterministic forecasting as a velocity component for stable probabilistic time-series predictions. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Covers forecasting uncertainty quantification, not deployment safety guardrails.
Deterministic Prediction as Global Velocity: Flow Matching for Probabilistic Time-Series Forecast The paper proposes a flow-matching model that integrates deterministic forecasting as a velocity component for stable probabilistic time-series predictions. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Addresses time-series forecasting architectures, not small or efficient LLMs.
Minimax rates for pointwise CATE estimation: the role of design smoothness The paper derives minimax risk bounds showing how covariate design smoothness affects pointwise conditional average treatment effect estimation. Preprint	Small and efficient models via arxiv · 15 Oct, 06:18	Not relevant Deals with theoretical statistics, not model efficiency or compression.
Minimax rates for pointwise CATE estimation: the role of design smoothness The paper derives minimax risk bounds showing how covariate design smoothness affects pointwise conditional average treatment effect estimation. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Theoretical causal bounds, not deployment safety or robustness.
Minimax rates for pointwise CATE estimation: the role of design smoothness The paper derives minimax risk bounds showing how covariate design smoothness affects pointwise conditional average treatment effect estimation. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Focuses on statistical causal estimation theory, not LLM benchmarks.
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	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Focuses on memory compression and quantization, not deployment safety or guardrails.

9,481–9,540 of 17,313← Previous Next →