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
State-Matched Sequence Conformal Prediction The paper proposes a sequential conformal prediction algorithm that groups calibration scores by state similarity rather than pure temporal recency. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on conformal prediction theory, not LLM benchmarking.
Digital Self-Interference Cancellation for In-Band Full-Duplex Systems Based on Transformer Neural Operators The paper applies transformer neural operators to model interference channels and cancel digital self-interference in full-duplex radio systems. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Neural operator for full-duplex radio, not general model compression.
Digital Self-Interference Cancellation for In-Band Full-Duplex Systems Based on Transformer Neural Operators The paper applies transformer neural operators to model interference channels and cancel digital self-interference in full-duplex radio systems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Radio frequency signal cancellation, not language model evaluation.
Digital Self-Interference Cancellation for In-Band Full-Duplex Systems Based on Transformer Neural Operators The paper applies transformer neural operators to model interference channels and cancel digital self-interference in full-duplex radio systems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Not related to AI safety, hallucinations, or model reliability.
SKG: A Law-Induced Engine from Inference to Induction The paper presents a mathematical framework unifying symplectic geometry, kernel methods, and physical conservation laws for physics-constrained learning. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Theoretical physics-informed gradient reasoning engine, not LLM benchmarks.
SKG: A Law-Induced Engine from Inference to Induction The paper presents a mathematical framework unifying symplectic geometry, kernel methods, and physical conservation laws for physics-constrained learning. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Geometric mechanical inference framework, not small language models or distillation.
SKG: A Law-Induced Engine from Inference to Induction The paper presents a mathematical framework unifying symplectic geometry, kernel methods, and physical conservation laws for physics-constrained learning. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address deployment safety, jailbreaks, or AI guardrails.
CktHGVAE: A Symmetric Design/State Heterogeneous Graph Variational Autoencoder for Transistor-level Analog/Mixed-signal Circuit Design and Analysis The paper introduces a heterogeneous graph variational autoencoder to model design and state variables jointly for analog circuit analysis. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Circuit design optimization, not small or efficient language models.
CktHGVAE: A Symmetric Design/State Heterogeneous Graph Variational Autoencoder for Transistor-level Analog/Mixed-signal Circuit Design and Analysis The paper introduces a heterogeneous graph variational autoencoder to model design and state variables jointly for analog circuit analysis. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Not concerned with AI safety, guardrails, or model hallucinations.
CktHGVAE: A Symmetric Design/State Heterogeneous Graph Variational Autoencoder for Transistor-level Analog/Mixed-signal Circuit Design and Analysis The paper introduces a heterogeneous graph variational autoencoder to model design and state variables jointly for analog circuit analysis. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Heterogeneous graph VAEs for analog circuits, not LLM evaluation.
High-Probability Regret Bounds for Feel-Good Thompson Sampling The paper establishes theoretical high-probability regret bounds for feel-good Thompson sampling algorithms in contextual bandit problems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Contextual bandit regret theory, not language model evaluation or benchmarks.
High-Probability Regret Bounds for Feel-Good Thompson Sampling The paper establishes theoretical high-probability regret bounds for feel-good Thompson sampling algorithms in contextual bandit problems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Not related to AI safety, red-teaming, or model deployment risks.
High-Probability Regret Bounds for Feel-Good Thompson Sampling The paper establishes theoretical high-probability regret bounds for feel-good Thompson sampling algorithms in contextual bandit problems. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Theoretical bandit algorithms, not model compression or efficient serving.
HankelKANet: Lightweight Path Loss Estimation via Physics-Inspired Hankel Convolution The paper proposes a neural network using physics-inspired Hankel convolution kernels for lightweight path loss estimation in wireless networks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address AI safety, hallucination, or deployment risks.
HankelKANet: Lightweight Path Loss Estimation via Physics-Inspired Hankel Convolution The paper proposes a neural network using physics-inspired Hankel convolution kernels for lightweight path loss estimation in wireless networks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on wireless path loss estimation, not LLM evaluation.
HankelKANet: Lightweight Path Loss Estimation via Physics-Inspired Hankel Convolution The paper proposes a neural network using physics-inspired Hankel convolution kernels for lightweight path loss estimation in wireless networks. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Specific wireless physics-informed network, not general small language models or compression.
Two-stage Residual Learning for Modelling Strength-Informed Optimisation of Thermal Systems The paper introduces a two-stage residual learning framework to detect local inaccuracies and guide the optimization of thermal systems. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on thermal systems optimization, not efficient model architectures.
Two-stage Residual Learning for Modelling Strength-Informed Optimisation of Thermal Systems The paper introduces a two-stage residual learning framework to detect local inaccuracies and guide the optimization of thermal systems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Not focused on AI safety, guardrails, or language model reliability.
Two-stage Residual Learning for Modelling Strength-Informed Optimisation of Thermal Systems The paper introduces a two-stage residual learning framework to detect local inaccuracies and guide the optimization of thermal systems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Surrogate modeling for thermal systems, not LLM evaluation or benchmarks.
Noisy-Label Calibration with Multiple Unreliable Labels The paper develops methods to evaluate and recalibrate binary predictors using noisy, multi-annotator labels without requiring clean ground truth. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model compression, distillation, or efficient inference.
Noisy-Label Calibration with Multiple Unreliable Labels The paper develops methods to evaluate and recalibrate binary predictors using noisy, multi-annotator labels without requiring clean ground truth. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Addresses general binary classification calibration under noisy labels, not LLM benchmarking.
Test-Time Forecasting: Recurrent Depth Enables In-Context Dynamics Prediction The paper uses recurrent gradient descent iterations during test time to enable foundation models to forecast unseen dynamical systems. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Addresses dynamical systems forecasting via recurrent depth, not general LLM compression.
Test-Time Forecasting: Recurrent Depth Enables In-Context Dynamics Prediction The paper uses recurrent gradient descent iterations during test time to enable foundation models to forecast unseen dynamical systems. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on dynamical systems forecasting architectures, not LLM evaluation.
Test-Time Forecasting: Recurrent Depth Enables In-Context Dynamics Prediction The paper uses recurrent gradient descent iterations during test time to enable foundation models to forecast unseen dynamical systems. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Not related to AI safety, hallucinations, or deployment guardrails.
Geometric Kernel Diversity: An Axiomatic Diversity Measure in Euclidean Space The paper introduces a continuous, monotonic geometric kernel measure and algorithms to efficiently compute diversity over Euclidean points. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Computational geometry for axiomatic diversity measures, not language model evaluation.
Geometric Kernel Diversity: An Axiomatic Diversity Measure in Euclidean Space The paper introduces a continuous, monotonic geometric kernel measure and algorithms to efficiently compute diversity over Euclidean points. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Theoretical diversity measure algorithms, not model efficiency or compression.
Geometric Kernel Diversity: An Axiomatic Diversity Measure in Euclidean Space The paper introduces a continuous, monotonic geometric kernel measure and algorithms to efficiently compute diversity over Euclidean points. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address AI safety, jailbreaks, or deployment guardrails.
On Inherent Privacy of Posterior Sampling: A Unified Rényi-Divergence Framework The paper develops a theoretical Rényi-divergence framework to measure inherent differential privacy guarantees obtained directly from Bayesian posterior sampling. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Addresses theoretical differential privacy bounds, not LLM evaluation frameworks.
On Inherent Privacy of Posterior Sampling: A Unified Rényi-Divergence Framework The paper develops a theoretical Rényi-divergence framework to measure inherent differential privacy guarantees obtained directly from Bayesian posterior sampling. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on Bayesian differential privacy theory, not efficient inference or SLMs.
Projected Wasserstein-Fisher-Rao Flows for Hawkes Background-Intensity Estimation The paper proposes a variational method using Wasserstein-Fisher-Rao geometry to estimate the background intensity of Hawkes processes. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on statistical manifold flows, not model compression or efficient serving.
Projected Wasserstein-Fisher-Rao Flows for Hawkes Background-Intensity Estimation The paper proposes a variational method using Wasserstein-Fisher-Rao geometry to estimate the background intensity of Hawkes processes. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Theoretical statistics for Hawkes processes, not LLM benchmarking or evaluation.
Projected Wasserstein-Fisher-Rao Flows for Hawkes Background-Intensity Estimation The paper proposes a variational method using Wasserstein-Fisher-Rao geometry to estimate the background intensity of Hawkes processes. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Not related to AI safety, hallucination, or deployment robustness.
InstructPro2: Natural Language and Biological Grounding Guided De Novo Protein Design The paper introduces a model for de novo protein sequence generation guided by natural language instructions and biological context encoders. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not study AI deployment safety, hallucination, guardrails, or red-teaming.
InstructPro2: Natural Language and Biological Grounding Guided De Novo Protein Design The paper introduces a model for de novo protein sequence generation guided by natural language instructions and biological context encoders. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address small language models, distillation, pruning, or efficient inference.
InstructPro2: Natural Language and Biological Grounding Guided De Novo Protein Design The paper introduces a model for de novo protein sequence generation guided by natural language instructions and biological context encoders. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on computational biology and protein sequence design, not LLM evaluation benchmarks.
Risk-Controlled DFT Selection for Machine-Learned Interatomic Potentials The paper applies conformal risk control to decide when to trust machine-learned interatomic potentials versus running expensive physics calculations. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Covers conformal risk control for interatomic potentials, not LLMs.
Risk-Controlled DFT Selection for Machine-Learned Interatomic Potentials The paper applies conformal risk control to decide when to trust machine-learned interatomic potentials versus running expensive physics calculations. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Materials discovery potential selection, not small model optimization.
Risk-Controlled DFT Selection for Machine-Learned Interatomic Potentials The paper applies conformal risk control to decide when to trust machine-learned interatomic potentials versus running expensive physics calculations. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Materials science simulation selection, not AI deployment safety.
Synthetic reaction data from transformation rules for graph-to-sequence pretraining in biodegradation pathway prediction The paper develops a graph-to-sequence pretraining framework using rule-generated synthetic reactions to predict chemical biodegradation pathways. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on biodegradation chemical pathways, not LLM benchmarks.
Synthetic reaction data from transformation rules for graph-to-sequence pretraining in biodegradation pathway prediction The paper develops a graph-to-sequence pretraining framework using rule-generated synthetic reactions to predict chemical biodegradation pathways. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Biochemical prediction framework, not model compression or efficiency.
Synthetic reaction data from transformation rules for graph-to-sequence pretraining in biodegradation pathway prediction The paper develops a graph-to-sequence pretraining framework using rule-generated synthetic reactions to predict chemical biodegradation pathways. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Chemical reaction modeling, not AI deployment safety.
On Global Joint Lower Bounds in Zeroth-Order Nonconvex Optimization The paper proves theoretical joint lower bounds on oracle complexity for zeroth-order nonconvex stochastic optimization. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Theoretical nonconvex optimization bounds, not LLM benchmarks.
On Global Joint Lower Bounds in Zeroth-Order Nonconvex Optimization The paper proves theoretical joint lower bounds on oracle complexity for zeroth-order nonconvex stochastic optimization. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Mathematical optimization complexity, not AI safety or alignment.
On Global Joint Lower Bounds in Zeroth-Order Nonconvex Optimization The paper proves theoretical joint lower bounds on oracle complexity for zeroth-order nonconvex stochastic optimization. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Purely theoretical optimization analysis, not model efficiency.
Annealed Lévy Diffusion via Gaussian Surrogate Reverse Dynamics The paper proposes a generative diffusion framework that anneals heavy-tailed noise into Gaussian noise to combine exploration with stable refinement. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Studies diffusion sampling mathematics, not LLM evaluation benchmarks.
Annealed Lévy Diffusion via Gaussian Surrogate Reverse Dynamics The paper proposes a generative diffusion framework that anneals heavy-tailed noise into Gaussian noise to combine exploration with stable refinement. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Deals with continuous diffusion processes, not model compression.
Annealed Lévy Diffusion via Gaussian Surrogate Reverse Dynamics The paper proposes a generative diffusion framework that anneals heavy-tailed noise into Gaussian noise to combine exploration with stable refinement. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Covers diffusion formulations, not LLM safety or guardrails.
ReGRA: Residual-Gradient Adaptation for Complementary Low-Rank Fine-Tuning The paper introduces a two-stage low-rank fine-tuning method that uses residual gradients to add complementary adaptation directions without expanding the rank. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Introduces a fine-tuning method, not a benchmarking suite.
ReGRA: Residual-Gradient Adaptation for Complementary Low-Rank Fine-Tuning The paper introduces a two-stage low-rank fine-tuning method that uses residual gradients to add complementary adaptation directions without expanding the rank. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on parameter-efficient adaptation, not deployment safety or guardrails.
TIMERSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models The paper presents an agentic system that iteratively writes and adjusts forecasting pipelines around frozen time-series models to improve predictions. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Covers time-series pipeline tuning, not AI safety or alignment.
TIMERSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models The paper presents an agentic system that iteratively writes and adjusts forecasting pipelines around frozen time-series models to improve predictions. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on automated time-series pipelines, not LLM evaluation benchmarks.
TIMERSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models The paper presents an agentic system that iteratively writes and adjusts forecasting pipelines around frozen time-series models to improve predictions. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Deals with agentic pipeline optimization, not model compression.
CAST: History-Adaptive Spectral Clipping for LLM Pretraining Introduces an adaptive spectral gradient clipping method to stabilize and accelerate large language model pretraining. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes a pretraining gradient clipping optimizer, not evaluation.
CAST: History-Adaptive Spectral Clipping for LLM Pretraining Introduces an adaptive spectral gradient clipping method to stabilize and accelerate large language model pretraining. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on pretraining optimization stability, not deployment safety.
Learning to Compute Is Not Learning to Retain or Reuse in Looped Transformers Investigates why looped transformers fail to retain or reuse earlier correct intermediate states when extended past training depth. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Investigates recurrent transformer mechanics, not LLM evaluation benchmarks.
Learning to Compute Is Not Learning to Retain or Reuse in Looped Transformers Investigates why looped transformers fail to retain or reuse earlier correct intermediate states when extended past training depth. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Explores mechanistic recurrent reasoning, not AI safety or guardrails.
When Does the Dependency Graph Help Code-to-Architecture Mapping? Evaluates whether dependency graphs improve automated mapping between source code files and system architectures. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Covers code architectural debt, not deployment safety or alignment.
When Does the Dependency Graph Help Code-to-Architecture Mapping? Evaluates whether dependency graphs improve automated mapping between source code files and system architectures. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Studies software dependency mapping rather than LLM evaluation benchmarks.
When Does the Dependency Graph Help Code-to-Architecture Mapping? Evaluates whether dependency graphs improve automated mapping between source code files and system architectures. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model compression, pruning, or efficient inference.
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.

2,581–2,640 of 13,754← Previous Next →