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
From Global to Tail Convergence: Dynamics of $α$-Stable Score-Based Generative Models Provides theoretical convergence bounds for score-based generative models driven by heavy-tailed stable processes. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Deals with theoretical score dynamics, not AI safety guardrails.
From Global to Tail Convergence: Dynamics of $α$-Stable Score-Based Generative Models Provides theoretical convergence bounds for score-based generative models driven by heavy-tailed stable processes. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Theoretical diffusion dynamics, not efficient serving or quantization.
SNACK: A Sequential Notation Framework for Probabilistic Graph Generation Proposes a sequential string notation to represent and generate valid graphs using standard sequence models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on graph generation notation, not LLM benchmarks.
SNACK: A Sequential Notation Framework for Probabilistic Graph Generation Proposes a sequential string notation to represent and generate valid graphs using standard sequence models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses molecular graph generation, not AI deployment safety.
SNACK: A Sequential Notation Framework for Probabilistic Graph Generation Proposes a sequential string notation to represent and generate valid graphs using standard sequence models. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Introduces string representation for graphs, not model compression.
Learning Who Benefits Most: Multi-Treatment Causal Ranking for Population Health Stratification Formulates clinical population health stratification as a multi-treatment causal ranking problem to prioritize interventions. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Healthcare causal ranking framework, not LLM benchmark evaluation.
Learning Who Benefits Most: Multi-Treatment Causal Ranking for Population Health Stratification Formulates clinical population health stratification as a multi-treatment causal ranking problem to prioritize interventions. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on clinical stratification, not AI safety or alignment.
Learning Who Benefits Most: Multi-Treatment Causal Ranking for Population Health Stratification Formulates clinical population health stratification as a multi-treatment causal ranking problem to prioritize interventions. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Deals with causal patient stratification, not model efficiency.
Reinforcement Learning via Advantage Guidance Uses supervised advantage conditioning to train reinforcement learning policies with improved stability. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Studies general RL policy extraction, not model compression techniques.
Reinforcement Learning via Advantage Guidance Uses supervised advantage conditioning to train reinforcement learning policies with improved stability. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant General RL policy extraction, not deployment safety or alignment.
Reinforcement Learning via Advantage Guidance Uses supervised advantage conditioning to train reinforcement learning policies with improved stability. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents a reinforcement learning algorithm, not LLM benchmarks.
Asymptotically Exact Negative Guidance of Diffusion Models via Positive-Unlabeled Learning Applies positive-unlabeled learning to establish mathematically exact negative guidance in diffusion generative models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Diffusion negative sampling theory, not deployment safety or guardrails.
Asymptotically Exact Negative Guidance of Diffusion Models via Positive-Unlabeled Learning Applies positive-unlabeled learning to establish mathematically exact negative guidance in diffusion generative models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Covers negative guidance in diffusion models, not LLM evaluation.
Asymptotically Exact Negative Guidance of Diffusion Models via Positive-Unlabeled Learning Applies positive-unlabeled learning to establish mathematically exact negative guidance in diffusion generative models. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Deals with diffusion sampling guidance, not small efficient models.
Efficient contextual partial monitoring with general function approximation Presents an efficient regression-based algorithm for contextual partial monitoring games with side information. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Theoretical bandit and game theory paper, not LLM evaluation.
Efficient contextual partial monitoring with general function approximation Presents an efficient regression-based algorithm for contextual partial monitoring games with side information. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Covers partial monitoring games, not model compression or efficiency.
Efficient contextual partial monitoring with general function approximation Presents an efficient regression-based algorithm for contextual partial monitoring games with side information. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on theoretical sequential learning, not deployment safety.
Calibration and Risk-Budgeted Decisions with System One Models Studies statistical calibration techniques to bound risk when fast language models make automated decisions under uncertainty. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model pruning, distillation, or inference optimization.
Calibration and Risk-Budgeted Decisions with System One Models Studies statistical calibration techniques to bound risk when fast language models make automated decisions under uncertainty. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on decision calibration and risk budgeting rather than benchmarks.
When Does Depth Help? Understanding Layerwise Inference in Tabular Foundation Models Analyzes layerwise inference across tabular foundation models to understand how depth and intermediate representations affect predictions. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Studies layer dynamics in tabular models without efficient serving methods.
When Does Depth Help? Understanding Layerwise Inference in Tabular Foundation Models Analyzes layerwise inference across tabular foundation models to understand how depth and intermediate representations affect predictions. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Analyzes internal representations of tabular models, not LLM benchmarking.
When Does Depth Help? Understanding Layerwise Inference in Tabular Foundation Models Analyzes layerwise inference across tabular foundation models to understand how depth and intermediate representations affect predictions. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Investigates tabular representations, not deployment safety or alignment.
JointServe: Serving Multi-Model Agents on Memory-Constrained GPUs Presents a serving system that dynamically manages GPU memory and prefetches weights for multi-model agent workflows. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on memory serving architecture rather than LLM evaluation benchmarks.
JointServe: Serving Multi-Model Agents on Memory-Constrained GPUs Presents a serving system that dynamically manages GPU memory and prefetches weights for multi-model agent workflows. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on GPU serving systems, not safety or guardrails.
CTIA-ODE: Continuous-Time Inter-Variable Attention for Long-Term Time Series Forecasting Introduces continuous-time inter-variable attention inside differential equation solvers for long-term multivariate time-series forecasting. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on time series ODE architectures, not LLM efficiency.
CTIA-ODE: Continuous-Time Inter-Variable Attention for Long-Term Time Series Forecasting Introduces continuous-time inter-variable attention inside differential equation solvers for long-term multivariate time-series forecasting. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Studies time series forecasting, not LLM safety, guardrails, or alignment.
CTIA-ODE: Continuous-Time Inter-Variable Attention for Long-Term Time Series Forecasting Introduces continuous-time inter-variable attention inside differential equation solvers for long-term multivariate time-series forecasting. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Addresses continuous-time time series forecasting, not LLM evaluation frameworks.
Rent-Seeking Language Models: Reward Models Set the Prices, Frequent Audits Curb the Rent Models reward overoptimization in reinforcement learning with human feedback as economic rent-seeking driven by proxy reward signals. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on RLHF reward gaming dynamics, not benchmark design or evaluation methods.
Rent-Seeking Language Models: Reward Models Set the Prices, Frequent Audits Curb the Rent Models reward overoptimization in reinforcement learning with human feedback as economic rent-seeking driven by proxy reward signals. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model compression, quantization, distillation, or serving efficiency.
Energy Consumption Analysis of Discrete Diffusion and Autoregressive Language Models Benchmarks the inference energy efficiency of diffusion language models compared to standard autoregressive architectures. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Analyzes energy consumption metrics rather than evaluating model task capabilities or benchmarks.
Energy Consumption Analysis of Discrete Diffusion and Autoregressive Language Models Benchmarks the inference energy efficiency of diffusion language models compared to standard autoregressive architectures. Peer-reviewed	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on inference hardware energy consumption, not deployment safety or robustness.
Text-Informed Training Data Filtering for Enhanced Virtual Flow Meter Performance Applies large language models to filter anomalies in industrial sensor training data using unstructured operational logs. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Deals with industrial sensor data cleaning, not AI deployment safety.
Text-Informed Training Data Filtering for Enhanced Virtual Flow Meter Performance Applies large language models to filter anomalies in industrial sensor training data using unstructured operational logs. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on industrial soft-sensor modeling, not efficient language models.
Text-Informed Training Data Filtering for Enhanced Virtual Flow Meter Performance Applies large language models to filter anomalies in industrial sensor training data using unstructured operational logs. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Uses LLMs for tabular data filtering, not LLM benchmark evaluation.
On the Optimisation Landscape of Overparameterised Weighted MMD Analyzes the optimization landscape of overparameterized weighted maximum mean discrepancy to provide finite-particle guarantees. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Purely mathematical analysis of maximum mean discrepancy optimisation landscapes.
On the Optimisation Landscape of Overparameterised Weighted MMD Analyzes the optimization landscape of overparameterized weighted maximum mean discrepancy to provide finite-particle guarantees. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Addresses theoretical kernel optimisation, not small or efficient models.
On the Optimisation Landscape of Overparameterised Weighted MMD Analyzes the optimization landscape of overparameterized weighted maximum mean discrepancy to provide finite-particle guarantees. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Mathematical optimisation theory without focus on deployment safety or reliability.
SPARS: A Stability-Plasticity-Aware Region Selection Framework for Continual Learning Mitigates catastrophic forgetting in continual learning through a hybrid approach combining parameter isolation and local regularization. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Addresses continual learning forgetting rather than model compression or inference efficiency.
SPARS: A Stability-Plasticity-Aware Region Selection Framework for Continual Learning Mitigates catastrophic forgetting in continual learning through a hybrid approach combining parameter isolation and local regularization. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on catastrophic forgetting in continual learning, not deployment safety.
SPARS: A Stability-Plasticity-Aware Region Selection Framework for Continual Learning Mitigates catastrophic forgetting in continual learning through a hybrid approach combining parameter isolation and local regularization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes a continual learning algorithm, not an LLM evaluation framework.
Occupancy Modeling via Drifting: One-Step Generation Enables Direct Policy Extraction Proposes a one-step generative approach to model occupancy for simplified policy extraction in reinforcement learning. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Addresses reinforcement learning policy extraction, not LLM benchmarking.
Occupancy Modeling via Drifting: One-Step Generation Enables Direct Policy Extraction Proposes a one-step generative approach to model occupancy for simplified policy extraction in reinforcement learning. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on RL occupancy models rather than small language models.
Occupancy Modeling via Drifting: One-Step Generation Enables Direct Policy Extraction Proposes a one-step generative approach to model occupancy for simplified policy extraction in reinforcement learning. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Deals with reinforcement learning sampling, not AI safety or guardrails.
Few-Shot Adaptation for Anomalous Sound Detection at Low False-Positive Rate Improves industrial sound anomaly detection by adapting unsupervised models with few fault samples to reduce false alarms. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Studies acoustic machinery fault detection, not language model evaluation.
Few-Shot Adaptation for Anomalous Sound Detection at Low False-Positive Rate Improves industrial sound anomaly detection by adapting unsupervised models with few fault samples to reduce false alarms. Peer-reviewed	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Addresses acoustic anomaly detection algorithms, not small language model efficiency.
Few-Shot Adaptation for Anomalous Sound Detection at Low False-Positive Rate Improves industrial sound anomaly detection by adapting unsupervised models with few fault samples to reduce false alarms. Peer-reviewed	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on industrial audio anomaly detection rather than AI system safety.
When Non-Adversarial Noise Hurts Learning: Benchmarking Skill Self-Evolution in Agents Introduces a benchmark to evaluate how language model agents learn and self-evolve reusable skills when exposed to non-adversarial noise. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on agent skill evolution and noise robustness, not model compression.
Beyond DFT electronic band structures via dual-space orbital learning Predicts high-level electronic band structures of materials using orbital information combined with low-cost baseline calculations. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Physics-informed material property prediction, not AI safety or jailbreak prevention.
Beyond DFT electronic band structures via dual-space orbital learning Predicts high-level electronic band structures of materials using orbital information combined with low-cost baseline calculations. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Application to material science and quantum mechanics, not LLM evaluation.
Beyond DFT electronic band structures via dual-space orbital learning Predicts high-level electronic band structures of materials using orbital information combined with low-cost baseline calculations. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Deals with electronic band structure prediction, not small language models.
Convergence Rates of Gaussian Process Thompson Sampling for Efficient Global Optimization Establishes theoretical convergence bounds for Gaussian process Thompson sampling under noiseless conditions in global optimization. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on Gaussian process sampling bounds, not efficient LLM serving.
Convergence Rates of Gaussian Process Thompson Sampling for Efficient Global Optimization Establishes theoretical convergence bounds for Gaussian process Thompson sampling under noiseless conditions in global optimization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Theoretical study of Bayesian optimization convergence rates, not LLM evaluation.
Convergence Rates of Gaussian Process Thompson Sampling for Efficient Global Optimization Establishes theoretical convergence bounds for Gaussian process Thompson sampling under noiseless conditions in global optimization. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses theoretical black-box optimization, not deployment safety or monitoring.
MorphingNet: Symbolic Regression Where the Network Is the Formula Treats symbolic regression as a continuous optimization problem using neural units that morph dynamically into mathematical formulas. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Studies continuous mathematical formula discovery rather than small language models.
MorphingNet: Symbolic Regression Where the Network Is the Formula Treats symbolic regression as a continuous optimization problem using neural units that morph dynamically into mathematical formulas. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents a neural architecture for symbolic regression, not LLM benchmarks.
MorphingNet: Symbolic Regression Where the Network Is the Formula Treats symbolic regression as a continuous optimization problem using neural units that morph dynamically into mathematical formulas. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on scientific formula recovery, not LLM safety or deployment risks.
LatentLoop: Scaling Looped MoE Models with Asymmetric Compute Allocation Allocates computation unevenly across passes in looped mixture-of-experts models to improve efficiency without losing performance. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on compute allocation efficiency rather than safety or reliability.
LatentLoop: Scaling Looped MoE Models with Asymmetric Compute Allocation Allocates computation unevenly across passes in looped mixture-of-experts models to improve efficiency without losing performance. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes an architectural design, not a language model evaluation method.
Minimax estimation of the pseudo-spectral gap of continuous Markov chains, with applications to empirical concentration bounds The paper develops a minimax-optimal estimator for the pseudo-spectral gap of continuous-state Markov chains from a single observed trajectory. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Examines theoretical concentration bounds rather than efficient model inference.
Minimax estimation of the pseudo-spectral gap of continuous Markov chains, with applications to empirical concentration bounds The paper develops a minimax-optimal estimator for the pseudo-spectral gap of continuous-state Markov chains from a single observed trajectory. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Theoretical statistics paper with no focus on AI deployment safety.

2,641–2,700 of 13,754← Previous Next →