Isolater - Feed

Ax William De Deyn, Michael Herty, Giovanni Samaey 7/1/2026

Mean-Field Model for Two-Layer Neural Networks Trained with Consensus-Based Optimization

Mean-field analysis of consensus-based optimization for training two-layer neural networks with comparison to Adam optimizer.

Ax Miao Zhang, Junpeng Li, Changchun Hua, Yana Yang 7/1/2026

A Unified and Stable Risk Minimization Framework for Weakly Supervised Learning with Theoretical Guarantees

Unified framework for weakly supervised learning across multiple annotation patterns with theoretical stability guarantees.

Ax Zo\"e Ruha Bell, Anvith Thudi, Olive Franzese-McLaughlin, Nicolas Papernot, Shafi Goldwasser 7/1/2026

Efficient Public Verification of Private ML via Regularization

Method for efficiently verifying differential privacy guarantees in machine learning models without retraining.

Ax Shobhita Sundaram, John Quan, Ariel Kwiatkowski, Kartik Ahuja, Yann Ollivier, Julia Kempe 7/1/2026

Teaching Models to Teach Themselves: Reasoning at the Edge of Learnability

SOAR framework using meta-RL for LLMs to generate automated curricula for self-improvement on low-success-rate reasoning tasks.

Ax Yiqi Su, Ray Lee, Jiaming Cui, Naren Ramakrishnan 7/1/2026

How (Not) to Hybridize Neural and Mechanistic Models for Epidemiological Forecasting

Analysis of hybrid approaches combining mechanistic epidemiological models with neural networks for disease forecasting under partial observability.

Ax Nicolas Anguita, Francesco Locatello, Andrew M. Saxe, Marco Mondelli, Flavia Mancini, Samuel Lippl, Clementine Domine 7/1/2026

A Theory of How Pretraining Shapes Inductive Bias in Fine-Tuning

Theoretical analysis of how pretraining shapes feature reuse and refinement during fine-tuning in neural networks.

Ax Chika Maduabuchi 7/1/2026

Entropy-Controlled Flow Matching

Entropy-controlled flow matching constrains information geometry in generative models to prevent semantic mode collapse.

Ax Xuan Yang, Jiayu Liu, Yuhang Lai, Hao Xu, Zhenya Huang, Ning Miao 7/1/2026

Step-Level Sparse Autoencoder for Reasoning Process Interpretation

Step-level sparse autoencoders for interpreting LLM chain-of-thought reasoning processes with fine-grained analysis.

Ax Xin Teng, Canyu Zhang, Shaoyi Zheng, Danyang Zhuo, Tianyi Zhou, Shenji Wan 7/1/2026

InfoFlow KV: Information-Flow-Aware KV Recomputation for Long Context

InfoFlow KV reduces long-context inference bottlenecks via information-flow-aware key-value cache recomputation.

Ax Yikai Gu, Lele Cao, Bo Zhao, Lei Lei, Lei You 7/1/2026

DISCOVER: A Solver for Distributional Counterfactual Explanations

Distributional counterfactual explanation method using optimal transport with statistical certification.

Ax Aravind Krishnan, Karolina Sta\'nczak, Dietrich Klakow 7/1/2026

On Optimizing Multimodal Jailbreaks for Spoken Language Models

Multimodal jailbreak attacks on spoken language models optimizing both text and audio simultaneously.

Ax Janne Perini, Rafael Bischof, Moab Arar, Ay\c{c}a Duran, Michael A. Kraus, Siddhartha Mishra, Bernd Bickel 7/1/2026

Pretrained Video Models as Differentiable Physics Simulators for Urban Wind Flows

WinDiNet uses pretrained video diffusion models as fast surrogate for urban wind flow CFD simulations.

Ax Zakaria Mhammedi, Alexander Rakhlin, Nneka Okolo 7/1/2026

End-to-End Efficient RL for Linear Bellman Complete MDPs with Deterministic Transitions

Computationally efficient RL algorithm with linear function approximation for MDPs with linear Bellman completeness.

Ax Weixin Chen, Antonio Vergari, Han Zhao 7/1/2026

Can VLMs Reason Robustly? A Neuro-Symbolic Investigation

Study of vision-language model reasoning robustness under distribution shifts using visual deductive reasoning tasks.

Ax Jawad Chowdhury, Ganesh Narasimha, Jan-Chi Yang, Yongtao Liu, Rama Vasudevan 7/1/2026

Quality-Controlled Active Learning via Gaussian Processes for Robust Structure-Property Learning in Autonomous Microscopy

Gaussian process active learning for autonomous microscopy to handle noisy experimental data in structure-property tasks.

Ax Huaiyang Wang, Xiaojie Li, Xiaohan Wang, Zhixia Zhang, Xiaodong Lu, Zixuan Huang, Jiajun Chai, Guojun Yin, Deqing Wang, Haoyi Zhou, Yaodong Yang, Jianxin Li, Yikun Ban 7/1/2026

Policy Improvement Reinforcement Learning

RL post-training method for LLMs/agents that explicitly verifies policy improvements over predecessors before updating.

Ax Tobias Schumacher, Simon Reichelt, Markus Strohmaier 7/1/2026

The Impact of Dimensionality on the Stability of Node Embeddings

Analysis of how embedding dimensionality affects training stability in node embedding methods across five algorithms.

Ax Feras Al Taha, Eilyan Bitar 7/1/2026

Wasserstein Distributionally Robust Risk-Sensitive Estimation via Conditional Value-at-Risk

Distributionally robust signal estimation using Wasserstein distance and conditional value-at-risk for risk-sensitive applications.

Ax Donghwan Lee 7/1/2026

Lyapunov-Certified Direct Switching Theory for Q-Learning

Switching linear system analysis framework for Q-learning convergence using joint spectral radius.

Ax Hanjun Cho, Gahyun Yoo, Hanseong Kim, Jay-Yoon Lee 7/1/2026

Generalizing Numerical Reasoning in Table Data through Operation Sketches and Self-Supervised Learning

TaNOS framework using operation sketches and self-supervised learning for robust numerical reasoning over domain-shifted tables.

Ax Minchan Kwon, Sunghyun Baek, Minseo Kim, Jaemyung Yu, Dongyoon Han, Junmo Kim 7/1/2026

Stable-GFlowNet: Toward Diverse and Robust LLM Red-Teaming via Contrastive Trajectory Balance

Stable-GFlowNet improves LLM red-teaming via contrastive trajectory balance to find diverse and robust adversarial attacks.

Ax Zhen Liu, Yuhan Liu, Jinjun Wang, Wei Song, Jianyi Liu, Jingwen Fu 7/1/2026

Structured Progressive Knowledge Activation for LLM-Driven Neural Architecture Search

LLM-driven NAS using structured knowledge activation to translate architectural priors into code edits while avoiding behavioral shifts.

Ax Ryan Lee, Jacob Biloki, Edward J. Hu, Jonathan May 7/1/2026

Sparse Layers are Critical to Scaling Looped Language Models

Sparse MoE layers enable looped transformer models to scale better than dense variants while reducing memory.

Ax Hongmin Li 7/1/2026

Targeted Tests for LLM Reasoning: An Audit-Constrained Protocol

Audit-constrained protocol for evaluating LLM reasoning robustness to semantically valid prompt variations.

Ax Chad Weatherly, Sen Lin 7/1/2026

Rethinking Continual Anomaly Detection on the Edge: Benchmarking Under Realistic Industrial Conditions

Benchmark for continual anomaly detection in industrial settings under realistic edge deployment constraints.

Ax Rupa Kurinchi-Vendhan, Sara Beery 7/1/2026

Finding Needles in the Haystack: Transductive Active Labeling in Ecology

Active learning evaluation protocol for ecological data labeling using transductive rather than inductive assessment.

Ax Siyu Chen, Miao Lu, Beining Wu, Heejune Sheen, Fengzhuo Zhang, Shuangning Li, Zhiyuan Li, Jose Blanchet, Tianhao Wang, Zhuoran Yang 7/1/2026

INFUSER: Influence-Guided Self-Evolution Improves Reasoning

INFUSER iterative co-training framework enables LLMs to self-improve reasoning with minimal external supervision using influence-guided generation.

Ax Jun Wen Leong 7/1/2026

Online Shift Detection and Conformal Adaptation for Deployed Safety Classifiers

Online monitoring system detects distributional shift in deployed safety classifiers using KS statistics and applies conformal adaptation.

Ax Tianyu Ding, Jianhong Xin, Juan Pablo De la Cruz Weinstein 7/1/2026

Keep Policy Gradient in Charge: Sibling-Guided Credit Distillation for Long-Horizon Tool-Use Agents

SGCD method improves long-horizon tool-use agents in RL by using sibling-guided credit distillation to better identify which actions lead to rewards.

Ax Silen Naihin, Lev Stambler 7/1/2026

Size Doesn't Matter: Cosine-Scored Sparse Autoencoders

Proposes replacing inner product scoring with cosine-based scoring in sparse autoencoders to prevent token norm features from claiming dictionary slots.

Ax Lukas Fesser, Hanlin Zhang, Michelle M. Li, Eric Wang, Bryan Perozzi, Shekoofeh Azizi, Sham M. Kakade, Marinka Zitnik 7/1/2026

How Post-Training Shapes Biological Reasoning Models

Studies how post-training stages in biological reasoning models affect performance and generalization across genomics, transcriptomics, and proteins.

Ax Soham Bhattacharjee, Dushyant Singh Chauhan, Salem Lahlou, Martin Takac, Nils Lukas 7/1/2026

Entropy-Gated Latent Recursion

Inference-time scaling method using layer-span recursion and entropy gating to improve language model reasoning without token sampling alone.

Ax Borui Cai, Yao Zhao 7/1/2026

Topological Neural Dynamics: A Neuron-wise Framework for Sequence Modeling

Neuron-wise sequence modeling framework allowing independent evolution of neurons rather than layer-wise shared dynamics.

Ax Rome Thorstenson 7/1/2026

Protein contacts are already in the attention: a single-forward-pass alternative to the Categorical Jacobian

Single-pass attention-based method for reading protein contacts from language models without per-residue perturbation.

Ax Jason Sulskis, Sathya Ravi 7/1/2026

Real vs. Complex Spectral Bases for Neural Operators: The Role of Green's Function Alignment

Hartley Neural Operator replacing complex FFT with real Discrete Hartley Transform for learning PDE solution operators.

Ax Kwok Chun Au, Adam Block 7/1/2026

Training for the Model You Return: Improving Optimization for Iterate-Averaged Language Models

Optimal control framework for improving language model training when returning iterate-averaged models instead of final iterations.

Ax DatologyAI, :, Matthew L. Leavitt, Siddharth Joshi, Haoli Yin, Rishabh Adiga, Haakon Mongstad, Alvin Deng, David Schwab, Bogdan Gaza, Ari Morcos 7/1/2026

Brevity is the Soul of Inference Efficiency: Inducing Concision in VLMs via Data Curation

Data curation approach for reducing inference cost in vision-language models by training models to generate concise outputs.

Ax Yijie Jin, Jiajun Xu, Yuxuan Liu, Chenkai Xu, Yi Tu, Jiajun Li, Dandan Tu, Xiaohui Yan, Kai Yu, Pengfei Liu, Zhijie Deng 7/1/2026

Multi-Block Diffusion Language Models

Multi-block diffusion language models enabling concurrent decoding of consecutive blocks for inter-block parallelism and flexible-length generation.

Ax Duc Anh Nguyen 7/1/2026

IG-Lens: Exact Additive Probability Attribution Across Transformer Layers via Telescoping Integrated Gradients

Method for exact additive probability attribution across transformer layers using telescoping integrated gradients for interpretability.

Ax Marek Polewczyk, Maximilian Schambach, Marco Spinaci, Sam Thelin, Johannes H\"ohne 7/1/2026

FlexTab: A Flexible Encoder-Decoder Architecture for In-Context Learning Across Diverse Tabular Tasks

Flexible encoder-decoder architecture for in-context learning on tabular data using task-agnostic embeddings and task-specific decoders.

Ax Liang Wang, Zhaoyang Xi, Zekai Xiang, Heng Meng, Qishan Zhang, Pingyi Zhou, Jin Liu, Litao Chen 7/1/2026

Arko-T: A Foundation Model for Text-to-Structured 3D Generation

4B-parameter foundation model generating executable parametric CAD programs from natural language text descriptions.

Ax Songxin Zhang, Zejian Xie, Zhuoyang Song, Cong lin, Junyu Lu, Jiaxing Zhang, Bingyi Jing 7/1/2026

HSAP: A Hierarchical Sequence-aware Parallelism for Hybrid-Context Generative Models

Hierarchical sequence parallelism framework for training large language models with packed sequences and hybrid-context generation.

Ax Kan Zhu, Mathew Jacob, Chenxi Ma, Yi Pan, Stephanie Wang, Arvind Krishnamurthy, Baris Kasikci 7/1/2026

TraceLab: Characterizing Coding Agent Workloads for LLM Serving

Characterizes coding agent workloads for LLM serving by collecting real day-to-day usage patterns across multiple agents and models.

Ax Kuang Huang, Dong Heon Cho, Boyuan Chen 7/1/2026

Automated Discovery of Operable Dynamics from Videos

Framework for automatically discovering low-dimensional operable representations of dynamical systems from videos without predefined state variables.

Ax Qirui Zheng, Xingbo Wang, Keyuan Cheng, Yunlong Lu, Muhammad Asif Ali, Lingfeng Li, Yongyi Wang, Wenxin Li 7/1/2026

From Multimodal Perception to Strategic Reasoning: A Survey on AI-Generated Game Commentary

Survey of AI-generated game commentary systems covering multimodal perception, NLG, and strategic reasoning for sports narration.

Ax Wooseok Ha, Yuansi Chen 7/1/2026

When few labeled target data suffice: a theory of semi-supervised domain adaptation via fine-tuning from multiple adaptive starts

Theoretical analysis of semi-supervised domain adaptation via fine-tuning with limited labeled target data under structural causal models.

Ax Savini Kashmira, Jayanaka Dantanarayana, Thamirawaran Sathiyalogeswaran, Krisztian Flautner, Lingjia Tang, Jason Mars 7/1/2026