Ax Aaditya L. Kachhadiya 5/14/2026

Local Inverse Geometry Can Be Amortized

Amortized learned surrogate (Deceptron) for solving nonlinear inverse problems by encoding curvature information into reusable reverse operator.

Ax Tien-Phat Nguyen, Truong Nguyen, Minh-Phuc Truong, Tuc Nguyen, James Bailey, Trung Le 5/14/2026

Spectral Flattening Is All Muon Needs: How Orthogonalization Controls Learning Rate and Convergence

Analysis of Muon optimizer showing spectral flattening mechanism enables larger learning rates and faster convergence through momentum orthogonalization.

Ax Chaokai Wu, Haofu Shi, Ningxuan Ma, Jianghong Ma, Xiaofeng Zhang 5/14/2026

MLGIB: Multi-Label Graph Information Bottleneck for Expressive and Robust Message Passing

Graph Information Bottleneck approach for multi-label graphs reducing over-squashing in deep message passing of graph neural networks.

Ax Changha Lee, Gyusang Cho 5/14/2026

ERPPO: Entropy Regularization-based Proximal Policy Optimization

Entropy regularization extension to multi-agent PPO addressing non-stationary observations in multi-dimensional cooperative reinforcement learning environments.

Ax Idan Barnea, Ofir Schlisselberg, Yishay Mansour 5/14/2026

Collaborating in Multi-Armed Bandits with Strategic Agents

Game-theoretic analysis of collaborative multi-agent bandit learning where strategic agents balance information sharing against free-riding incentives.

Ax Huiqi Deng, Yibo Li, Quanshi Zhang, Peng Zhang, Hongbin Pei, Xia Hu 5/14/2026

Understanding Generalization through Decision Pattern Shift

Decision Pattern Shift framework analyzing how deep neural network internal decision mechanisms evolve from training to test data for understanding generalization failures.

Ax Hung Le, Svetha Venkatesh 5/14/2026

Continual Fine-Tuning of Large Language Models via Program Memory

Parameter-efficient fine-tuning method for LLMs using program memory to balance rapid adaptation and knowledge retention in continual learning settings.

Ax Maxwell Standen, Junae Kim, Claudia Szabo 5/14/2026

Finding the Weakest Link: Adversarial Attack against Multi-Agent Communications

Adversarial attack methods against multi-agent reinforcement learning systems using Jacobian-based gradient information to identify vulnerable communications.

Ax Han Wang, Ying Wang, Bing Wang 5/14/2026

A Hybrid Tucker-LSTM Tensor Network Model for SOC Prediction in Electric Vehicles

arXiv paper: Hybrid Tucker-LSTM tensor network for battery state-of-charge prediction in electric vehicles.

Ax Stefan Stojanovic, Alexandre Proutiere 5/14/2026

Switching Successor Measures for Hierarchical Zero-shot Reinforcement Learning

arXiv paper: Hierarchical reinforcement learning with switching successor measures for zero-shot RL without fixed temporal abstractions.

Ax Paul Jeha, Anastasiia Sedova, Louis B\'ethune, Skyler Seto, Jes Frellsen, Pierre Ablin, Natalie Schluter 5/14/2026

Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings

arXiv paper: Bilingual pre-training outperforms hyperparameter tuning in data-constrained low-resource language settings.

Ax Xinyu Liu, Kechen Jiao, Chunyang Xiao, Runsong Zhao, Junhao Ruan, Bei Li, Jiahao Liu, Qifan Wang, Xin Chen, Jingang Wang, Tong Xiao, JingBo Zhu 5/14/2026

Teacher-Guided Policy Optimization for LLM Distillation

arXiv paper: Teacher-Guided Policy Optimization for LLM distillation using Reverse KL to improve student-teacher convergence.

Ax Linghao Jin, Chufan Shi, Huijuan Wang, Nuan Wen, Zhengzhong Liu, Eric Xing, Xuezhe Ma 5/14/2026

EMO: Frustratingly Easy Progressive Training of Extendable MoE

arXiv paper: EMO framework for progressive training of Mixture-of-Experts models addressing memory and communication efficiency bottlenecks.

Ax Yuka Hashimoto, Tomoharu Iwata 5/14/2026

Unified generalization analysis for physics informed neural networks

arXiv paper: Generalization bounds analysis for Physics-Informed Neural Networks (PINNs) and variational variants.

Ax Deepak Warrier, Raja Sekhar Pappala 5/14/2026

Chem-GMNet: A Sphere-Native Geometric Transformer for Molecular Property Prediction

arXiv paper: Chem-GMNet geometric transformer for molecular property prediction using domain-native architecture over generic SMILES models.

Ax Mert Cihangiroglu, Alessandro Pegoraro, Phillip Rieger, Antonino Nocera, Ahmad-Reza Sadeghi 5/14/2026

LightSplit: Practical Privacy-Preserving Split Learning via Orthogonal Projections

LightSplit privacy-preserving split learning using orthogonal projections to reduce communication overhead and prevent reconstruction attacks.

Ax Ian Osband 5/14/2026

Delightful Exploration

Exploration algorithm balancing uncertainty resolution with action budget using expected improvement and surprisal gating.

Ax Marco Angioli, Kevin Johansson, Antonello Rosato, Amy Loutfi, Denis Kleyko 5/14/2026

Contextual Bandits for Resource-Constrained Devices using Probabilistic Learning

Contextual bandits algorithm optimized for resource-constrained devices using probabilistic learning.

Ax Coleman Hooper, Minwoo Kang, Suhong Moon, Nicholas Lee, Eric Wen, John Wawrzynek, Michael W. Mahoney, Yakun Sophia Shao, Amir Gholami, Kurt Keutzer 5/14/2026

Building Interactive Real-Time Agents with Asynchronous I/O and Speculative Tool Calling

Real-time AI agents using asynchronous I/O and speculative tool calling for sub-1-second latency in interactive applications.

Ax Sungwoo Goo, Hwi-yeol Yun, Sangkeun Jung 5/14/2026

Phasor Memory Networks: Stable Backpropagation Through Time for Scalable Explicit Memory

Phasor Memory Networks architecture enabling stable backpropagation for explicit memory in language models via unitary dynamics.

Ax Eden Nagar, Ya-Wei Eileen Lin, Ron Levie 5/14/2026

Beyond Oversquashing: Understanding Signal Propagation in GNNs Via Observables

Analysis of signal propagation in GNNs addressing information loss through oversmoothing and oversquashing phenomena.

Ax Jan Arne Telle, Brigt H{\aa}vardstun, Jose Hernandez-Orallo 5/14/2026

Teaching and Learning under Deductive Errors

Teaching framework for machine learning that accounts for deductive errors in learners like LLMs during few-shot learning.

Ax Lilin Zhang, Yimo Guo, Yue Li, Jiancheng Shi, Xianggen Liu 5/14/2026

Taming the Long Tail: Rebalancing Adversarial Training via Adaptive Perturbation

Adversarial training approach addressing long-tail data imbalance through adaptive perturbations and theoretical analysis.

Ax Akhil Premkumar, Sarah Lucioni 5/14/2026

The Diffusion Encoder

Novel encoder architecture replacing VAE encoders with diffusion models for improved latent representation learning.

Ax Tobias Schm\"ahling, Matthias Burkhardt, Tobias Windisch 5/14/2026

Trajectory-Level Data Augmentation for Offline Reinforcement Learning

Data augmentation method for offline reinforcement learning using trajectory-based techniques to train models from limited suboptimal data.

Ax Neeratyoy Mallik, Maciej Janowski, Johannes Hog, Herilalaina Rakotoarison, Josif Grabocka, Frank Hutter, Aaron Klein 5/14/2026

When is Warmstarting Effective for Scaling Language Models?

Research on warmstarting techniques for scaling language models, analyzing initialization constraints and growth strategies for training efficiency.

Ax Marc Molina Van den Bosch, Riccardo Taiello, Albert Sund Aillet, Andrea Protani, Miguel Angel Gonzalez Ballester, Luigi Serio 5/14/2026

DP-KFC: Data-Free Preconditioning for Privacy-Preserving Deep Learning

Data-free second-order preconditioning for differentially private deep learning without privacy budget consumption.

Ax Divij Khaitan, Ashish Tiwari 5/14/2026

LIFT: Last-Mile Fine-Tuning for Table Explicitation

Pipeline combining pretrained LLM table extraction with fine-tuned small model error repair for low data requirements.

Ax Ammar Mahran, Artavazd Maranjyan, Peter Richt\'arik 5/14/2026

Rescaled Asynchronous SGD: Optimal Distributed Optimization under Data and System Heterogeneity

Asynchronous SGD with gradient rescaling for distributed optimization under data and system heterogeneity.

Ax JaeHyeok Doo, Byeongguk Jeon, Seonghyeon Ye, Kimin Lee, Minjoon Seo 5/14/2026

Q-Flow: Stable and Expressive Reinforcement Learning with Flow-Based Policy

Flow-based policy for stable and expressive reinforcement learning without backpropagation through solvers.

Ax Arkady Gonoskov 5/14/2026

Twincher: Bijective Representation Learning for Robust Inversion of Continuous Systems

Bijective representation learning framework for robust inversion of continuous forward processes.

Ax Chenyu Zhou, Hongpei Li, Yuerou Liu, Jianghao Lin, Dongdong Ge, Yinyu Ye 5/14/2026

OSDN: Improving Delta Rule with Provable Online Preconditioning in Linear Attention

Improved delta rule with online preconditioning for linear attention in state-space models.

Ax Katarzyna Kobalczyk, Mihaela van der Schaar 5/14/2026

Discovery of Hidden Miscalibration Regimes

Method for discovering hidden miscalibration in model confidence across different input types.

Ax Amirmehdi Jafari Fesharaki, Mohammadamin Rami, Aslan Tchamkerten 5/14/2026

Effective Context in Transformers: An Analysis of Fragmentation and Tokenization

Analysis of how tokenization and representation choices affect transformer context window effectiveness and information exposure.

Ax Qingyun Zou, Feng Yu, Hongshi Tan, Yao Chen, Bingsheng He, WengFai Wong 5/14/2026

HLS-Seek: QoR-Aware Code Generation for High-Level Synthesis via Proxy Comparative Reward Reinforcement Learning

RL-based LLM approach for high-level synthesis code generation using comparative rewards for quality optimization.

Ax Ye Wang, Jing Liu, Toshiaki Koike-Akino 5/14/2026

Temper and Tilt Lead to SLOP: Reward Hacking Mitigation with Inference-Time Alignment

Inference-time alignment technique using temperature adjustment to mitigate reward hacking in LLM outputs.

Ax Haonan Yuan, Qingyun Sun, Junhua Shi, Xingcheng Fu, Jianxin Li, Philip S. Yu 5/14/2026

Decoupled and Divergence-Conditioned Prompt for Multi-domain Dynamic Graph Foundation Models

Foundation model for dynamic graphs across multiple domains using decoupled prompts for multi-domain pretraining.

Ax Asim Osman, Sasha Abramowitz, Mark Bergh, Ulrich Armel Mbou Sob, Ruan John de Kock, Omayma Mahjoub, Oussama Hidaoui, Noah De Nicola, Arnol Manuel Fokam, Felix Chalumeau, Daniel Rajaonarivonivelomanantsoa, Siddarth Singh, Refiloe Shabe, Juan Claude Formanek, Simon Verster Du Toit, Arnu Pretorius 5/14/2026

Self-Supervised On-Policy Reinforcement Learning via Contrastive Proximal Policy Optimisation

Self-supervised contrastive reinforcement learning algorithm for discrete action spaces without hand-crafted rewards.

Ax Yatin Dandi, Matteo Vilucchio, Luca Arnaboldi, Hugo Tabanelli, Florent Krzakala 5/14/2026

Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning

Neural Low-Degree Filtering (Neural LoFi) provides spectral theory of hierarchical feature learning in gradient-based deep learning.

Ax Ishaq Hamza, Zaiwei Chen 5/14/2026

Achieving $\epsilon^{-2}$ Sample Complexity for Single-Loop Actor-Critic under Minimal Assumptions

Proves first Õ(ε⁻²) sample complexity for single-loop off-policy actor-critic methods under minimal assumptions.

Ax Yang Bai, Kaiyuan Liu, Ziyuan Zhuang, Jiahong Zhou, Rongxiang Weng, Xin Chen, Jingang Wang, Xunliang Cai 5/14/2026

Multi-Objective and Mixed-Reward Reinforcement Learning via Reward-Decorrelated Policy Optimization

Reward-Decorrelated Policy Optimization (RDPO) for multi-task and mixed-reward reinforcement learning with heterogeneous rewards.

Ax Namrata Shivagunde, Vijeta Deshpande, Sherin Muckatira, Anna Rumshisky 5/14/2026

Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training

Geometric and spectral study of low-rank pre-training for LLMs, comparing generalization to full-rank training beyond perplexity.

Ax Zhenan Yu, Guangxin Jiang, Jin Yang 5/14/2026

Three-Stage Learning Unlocks Strong Performance in Simple Models for Long-Term Time Series Forecasting

Three-stage learning approach for long-term time series forecasting using simple linear models and MLPs without complex architectures.

Ax Iskander Azangulov, Leo Zhang 5/14/2026

Sampling from Flow Language Models via Marginal-Conditioned Bridges

Proposes sampling method for flow language models using marginal-conditioned bridges to preserve posterior marginal structure.

Ax Shashaank Aiyer, Yishay Mansour, Shay Moran, Han Shao, Tom Waknine 5/14/2026

Scale-Sensitive Shattering: Learnability and Evaluability at Optimal Scale

Establishes scale-sensitive generalization of PAC learning fundamental theorem using fat-shattering dimension at optimal scales.

Ax Jason Gaitonde, Frederic Koehler, Elchanan Mossel, Joonhyung Shin, Allan Sly 5/14/2026

A Hierarchical Language Model with Predictable Scaling Laws and Provable Benefits of Reasoning

Analyzes hierarchical synthetic languages with exact k-gram ansatz to derive explicit scaling laws and benefits of reasoning in transformers.

Ax Yuetai Li, Fengqing Jiang, Yichen Feng, Kaiyuan Zheng, Luyao Niu, Bhaskar Ramasubramanian, Basel Alomair, Linda Bushnell, Radha Poovendran 5/14/2026

Polyhedral Instability Governs Regret in Online Learning

Analyzes regret in online learning over combinatorial actions using convex relaxations, governed by polyhedral instability.

Ax Hsing-Huan Chung, Shijun Li, Yoav Wald, Xing Han, Suchi Saria, Joydeep Ghosh 5/14/2026

MILM: Large Language Models for Multimodal Irregular Time Series with Informative Sampling

MILM extends large language models to handle multimodal irregular time series with asynchronous observations and textual channels.

Ax Amer Essakine, Claire Vernade 5/14/2026

Tight Sample Complexity Bounds for Entropic Best Policy Identification

Derives tight sample complexity bounds for best-policy identification in risk-sensitive reinforcement learning with entropic risk measures.

Ax Valentin Six, Frederik Panse, Mathis Fajeau, Lancelot Da Costa, Mridul Sharma, Alfonso Amayuelas, Tim Z. Xiao, David Hyland, Philipp Hennig, Bernhard Sch\"olkopf 5/14/2026

Learning POMDP World Models from Observations with Language-Model Priors

Learns POMDP world models from observation-action trajectories using language model priors for agents in partially observable environments.