Ax Wenhao Dai, Haodong Deng, Mengfei Rong, Xinyu Yang, Hongyu Liu, Fangxin Liu, Hailong Yang, Qianwen Cao, Qingxiao Sun 5/20/2026

Accelerating Sparse Transformer Inference on GPU

Optimizes sparse Transformer inference on GPUs through parallelization and operator fusion to reduce computation from mask-induced sparsity in LLMs.

Ax Fei Liu, Rui Zhang, Xi Lin, Zhichao Lu, Qingfu Zhang 5/20/2026

Fine-tuning Large Language Model for Automated Algorithm Design

Fine-tuning approach for LLMs specialized in algorithm design through supervised learning, comparing against general-purpose models.

Ax Carlos Barroso-Fern\'andez, Alejandro Calvillo-Fernandez, Antonio de la Oliva, Carlos J. Bernardos 5/20/2026

FAWN: A MultiEncoder Fusion-Attention Wave Network for Integrated Sensing and Communication Indoor Scene Inference

Fusion-attention network for indoor scene inference in integrated sensing and communication systems using multi-encoder approach.

Ax Hao Wu, Yuan Gao, Fan Xu, Fan Zhang, Qingsong Wen, Kun Wang, Xiaomeng Huang, Xian Wu 5/20/2026

Differential-Integral Neural Operator for Long-Term Turbulence Forecasting

Neural operator method for long-term turbulence forecasting combining differential and integral operations for physical fidelity.

Ax Rui Ai, Yuqi Pan, David Simchi-Levi, Milind Tambe, Haifeng Xu 5/20/2026

Beyond Majority Voting: LLM Aggregation by Leveraging Higher-Order Information

Aggregation algorithms for multi-agent LLM reasoning leveraging model heterogeneity and answer correlations beyond majority voting.

Ax Ankit Bhardwaj, Ananth Balashankar, Lakshminarayanan Subramanian 5/20/2026

FieldFormer: Locality-Aware Transformers for Spatio-Temporal Modeling on Sparse Sensor Networks

Transformer architecture for spatio-temporal field reconstruction from sparse sensor networks with locality inductive biases.

Ax Wei Guo, Jaemoo Choi, Yuchen Zhu, Molei Tao, Yongxin Chen 5/20/2026

Proximal Diffusion Neural Sampler

Framework for training diffusion-based neural samplers on multimodal distributions using proximal stochastic optimal control.

Ax Benedikt Franke, Florian Heinrich, Markus Lange, Arne Raulf 5/20/2026

Robustness and Regularization in Hierarchical Re-Basin

Hierarchical approach to model merging via Re-Basin algorithm that improves robustness properties of merged neural networks.

Ax Sanjari Srivastava, Gang Li, Cheng Chang, Rishu Garg, Manpreet Kaur, Charlene Y. Lee, Yuezhang Li, Yining Mao, Ignacio Cases, Yanan Xie, Peng Qi 5/20/2026

WARC-Bench: Web Archive Based Benchmark for GUI Subtask Executions

Benchmark with 438 web navigation subtasks for evaluating multimodal AI agents on real-world UI interactions.

Ax Simon Jaxy, Anton Theys, Patrick Willett, W. Chris Carleton, Ralf Vandam, Pieter Libin 5/20/2026

Needles in the Landscape: Semi-Supervised Pseudolabeling for Archaeological Site Discovery under Label Scarcity

Semi-supervised deep learning method for archaeological site discovery using asymmetric pseudolabeling under label scarcity.

Ax Hongliang Lu, Yuhang Wen, Pengyu Cheng, Ruijin Ding, Jiaqi Guo, Haotian Xu, Chutian Wang, Haonan Chen, Xiaoxi Jiang, Guanjun Jiang 5/20/2026

Search Self-play: Pushing the Frontier of Agent Capability without Supervision

RL method for training LLM agents without hand-crafted tasks via self-play search, enabling unsupervised agent capability scaling.

Ax Omar El Mansouri, Fathinah Asma Izzati, Mohamed El Amine Seddik, Salem Lahlou 5/20/2026

Noise-corrected GRPO: From Noisy Rewards to Unbiased Gradients

Noise-robust variant of Group Relative Policy Optimization for LLM alignment and reasoning models under reward noise.

Ax Jie Peng, Rui Wang, Qiang Wang, Zhewei Wei, Bin Tong, Guan Wang, Bo Zheng 5/20/2026

Beyond Leakage and Complexity: Towards Realistic and Efficient Information Cascade Prediction

Method addressing temporal leakage and data limitations in information cascade prediction for social networks.

Ax Parvin Nazari, Bojian Hou, Davoud Ataee Tarzanagh, Li Shen, George Michailidis 5/20/2026

Stochastic Regret Guarantees for Online Zeroth- and First-Order Bilevel Optimization

Theoretical framework for online bilevel optimization with stochastic regret guarantees for dynamic ML objectives.

Ax Lin Guan, Jia-Qi Yang, Zhishan Zhao, Beichuan Zhang, Bo Sun, Xuanyuan Luo, Jinan Ni, Xiaowen Li, Yuhang Qi, Zhifang Fan, Hangyu Wang, Qiwei Chen, Yi Cheng, Feng Zhang, Xiao Yang 5/20/2026

Make It Long, Keep It Fast: End-to-End 10K Long User Behavior Sequence Modeling for Billion-Scale Douyin Recommendation

Industrial recommendation system scaling long-sequence user behavior modeling to 10K histories using cross-attention mechanisms.

Ax Aihua Zhu, Rui Su, Qinglin Zhao, Li Feng, Meng Shen, Shibo He 5/20/2026

Hierarchical Schedule Optimization for Fast and Robust Diffusion Model Sampling

Training-free optimization technique for accelerating diffusion model sampling by finding optimal timestep schedules.

Ax Houtan Ghaffari, Lukas Rauch, Paul Devos 5/20/2026

Data-Efficient Self-Supervised Algorithms for Fine-Grained Birdsong Analysis

Self-supervised ML method for annotating birdsong audio with minimal training data using residual neural networks.

Ax Yoonhyuk Choi, Jiho Choi, Jiwoo Kang 5/20/2026

Gauge-Equivariant Graph Networks via Self-Interference Cancellation

Graph neural network architecture using gauge equivariance and interference cancellation for heterophilic graph learning.

Ax Shravan Chaudhari, Yoav Wald, Suchi Saria 5/20/2026

Open-Set Domain Adaptation Under Background Distribution Shift: Challenges and A Provably Efficient Solution

Research on domain adaptation for ML systems handling data distribution shifts and open-set recognition with theoretical guarantees.

Ax Nitin Priyadarshini Shankar, Vaibhav Singh, Sheetal Kalyani, Christian Maciocco 5/20/2026

BERTO: Intent-Driven Network Time Series Forecasting via Natural Language Operator Preferences

BERTO: BERT-based framework for network traffic forecasting with natural language preference operators.

Ax Sara Patel, Mingxun Zhou, Giulia Fanti 5/20/2026

MaxShapley: Towards Incentive-compatible Generative Search with Fair Context Attribution

MaxShapley algorithm for fair attribution and compensation of content sources in generative search systems.

Ax Zhuo Li, Pengyu Cheng, Zhechao Yu, Feifei Tong, Anningzhe Gao, Tsung-Hui Chang, Xiang Wan, Erchao Zhao, Xiaoxi Jiang, Guanjun Jiang 5/20/2026

Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidance

Information-theoretic approach to reduce inductive bias in reward models for RLHF training of LLMs.

Ax Qianyi Chen, Bo Li 5/20/2026

Colorful Pinball: Density-Weighted Quantile Regression for Conditional Guarantee of Conformal Prediction

Conformal prediction method using density-weighted quantile regression for conditional coverage guarantees in ML.

Ax Yuexuan Xia, Yinghao Zhang, Yalin Liu, Hong-Ning Dai, Yong Xia 5/20/2026

FedBiCross: A Bi-Level Optimization Framework to Tackle Non-IID Challenges in Data-Free One-Shot Federated Learning on Medical Data

Federated learning framework using bi-level optimization for one-shot knowledge distillation on non-IID medical data.

Ax Junyi Liao, Zihan Zhu, Ethan Fang, Zhuoran Yang, Vahid Tarokh 5/20/2026

Decoding Rewards in Competitive Games: Inverse Game Theory with Entropy Regularization

Framework for recovering reward functions in two-player games using inverse game theory with entropy regularization.

Ax Qiyang Li, Sergey Levine 5/20/2026

Q-learning with Adjoint Matching

Q-learning algorithm for continuous-action RL that optimizes diffusion/flow-matching policies using first-order critic information.

Ax Song Xia, Meiwen Ding, Chenqi Kong, Wenhan Yang, Xudong Jiang 5/20/2026

Feature-Space Smoothing: Certified Robustness of Deep Representations

Defense framework providing certified robustness for deep learning models against feature-space attacks.

Ax Xiuyu Li, Jinkai Zhang, Mingyang Yi, Yu Li, Longqiang Wang, Yue Wang, Ju Fan 5/20/2026

ETS: Energy-Guided Test-Time Scaling for Training-Free RL Alignment

Training-free inference method for RL-aligned LLMs using energy-guided sampling without costly post-training.

Ax Khiem Le, Phuc Nguyen, Youssef Mroueh, Chi-Heng Lin, Shangqian Gao, Ting Hua, Nitesh V. Chawla 5/20/2026

Transformation-Augmented GRPO for Enhancing Exploration in Reasoning of Large Language Models

Improves GRPO reinforcement learning for LLMs by addressing gradient vanishing and diversity collapse through transformation augmentation.

Ax Erfan Miahi, Eugene Belilovsky 5/20/2026

Understanding and Exploiting Weight Update Sparsity for Communication-Efficient Distributed RL

Analysis of weight update sparsity in distributed RL for LLM post-training, showing 99% invisible updates enabling communication-efficient training.

Ax Pourya Shamsolmoali, Masoumeh Zareapoor 5/20/2026

Finding Structure in Continual Learning

Continual learning method using Douglas-Rachford operator to reduce gradient conflicts without external memory or parameter regularization.

Ax Jaemoo Choi, Yuchen Zhu, Wei Guo, Petr Molodyk, Bo Yuan, Jinbin Bai, Yi Xin, Molei Tao, Yongxin Chen 5/20/2026

Rethinking the Design Space of Reinforcement Learning for Diffusion Models: On the Importance of Likelihood Estimation Beyond Loss Design

Analysis of RL for diffusion models showing importance of likelihood estimation beyond loss design for policy-gradient methods on visual tasks.

Ax Yu-Ang Lee, Ching-Yun Ko, Pin-Yu Chen, Mi-Yen Yeh 5/20/2026

Learning Rate Matters: Vanilla LoRA May Suffice for LLM Fine-tuning

Study demonstrating vanilla LoRA with proper learning rate tuning matches or exceeds recent LLM fine-tuning variants, challenging complexity claims.

Ax Riccardo Barbano, Alexander Denker, Zeljko Kereta, Runchang Li, Francisco Vargas 5/20/2026

CMAD: Cooperative Multi-Agent Diffusion via Stochastic Optimal Control

CMAD framework for composing multiple pre-trained diffusion models via stochastic optimal control without explicit target distribution knowledge.

Ax Michal Nauman, Marek Cygan, Pieter Abbeel 5/20/2026

Reward-Conditioned Reinforcement Learning

Reward-Conditioned RL (RCRL) off-policy method conditioning agents on reward parameterizations for robustness to reward misspecification.

Ax Ziwei Wang, Zhentao He, Xingyi He, Hongbin Wang, Tianwang Jia, Jingwei Luo, Siyang Li, Xiaoqing Chen, Dongrui Wu 5/20/2026

Synthetic Data Generation for Brain-Computer Interfaces: Overview, Benchmarking, and Future Directions

Overview and benchmarking of synthetic data generation methods for brain-computer interfaces to address limited neural recording data.

Ax Yifan Zhang, Liang Zheng 5/20/2026

RE-SAC: Disentangling aleatoric and epistemic risks in bus fleet control: A stable and robust ensemble DRL approach

RE-SAC ensemble DRL approach for bus fleet control that disentangles aleatoric and epistemic uncertainty to improve stability.

Ax Dharshan Kumaran, Nathaniel Daw, Simon Osindero, Petar Veli\v{c}kovi\'c, Viorica Patraucean 5/20/2026

Causal Evidence that Language Models use Confidence to Drive Behavior

Causal evidence that language models use internal confidence signals to control behavior like answer abstraction. Four-phase investigation paradigm.

Ax Manuel Milling, Andreas Triantafyllopoulos, Alexander Gebhard, Simon Rampp, Bj\"orn W. Schuller 5/20/2026

How Class Ontology and Data Scale Affect Audio Transfer Learning

Empirical study examining how class ontology and data scale affect transfer learning performance in audio deep learning tasks.

Ax Leonard S. Pleiss, James Harrison, Maximilian Schiffer 5/20/2026

Target-Aligned Reinforcement Learning

Target-Aligned Reinforcement Learning (TARL) addresses stability-recency tradeoff in value-based deep RL by refining target network mechanisms.

Ax Yuyuan Chen, Shiyi Wang, Peter Potaptchik, Jaeyeon Kim, Michael S. Albergo 5/20/2026

Discrete Tilt Matching

Discrete Tilt Matching: Likelihood-free RL fine-tuning for masked diffusion LLMs. Addresses intractability of sequence-level likelihoods in masked language model training.

Ax Yusuf Kesmen, Fay Elhassan, Jiayi Ma, Julien Stalhandske, Yena Chang, David Sasu, Alexandra Kulinkina, Akhil Arora, Lars Klein, Mary-Anne Hartley 5/20/2026

MoBayes: A Modular Bayesian Framework for Separating Reasoning from Language in Conversational Clinical Decision Support

MoBayes: Modular Bayesian framework separating reasoning from language in clinical decision support. Adds explicit posterior tracking and auditable reasoning to LLM-based systems.

Ax Jiachen Liu, Jiaxin Pei, Jintao Huang, Chenglei Si, Ao Qu, Xiangru Tang, Runyu Lu, Lichang Chen, Xiaoyan Bai, Haizhong Zheng, Carl Chen, Zhiyang Chen, Haojie Ye, Yujuan Fu, Zexue He, Zijian Jin, Zhenyu Zhang, Shangquan Sun, Maestro Harmon, John Dianzhuo Wang, Jianqiao Zeng, Jiachen Sun, Mingyuan Wu, Baoyu Zhou, Chenyu You, Shijian Lu, Yiming Qiu, Fan Lai, Yuan Yuan, Yao Li, Junyuan Hong, Ruihao Zhu, Beidi Chen, Alex Pentland, Ang Chen, Mosharaf Chowdhury, Zechen Zhang 5/20/2026

The Last Human-Written Paper: Agent-Native Research Artifacts

Proposes agent-native research artifacts to replace linear paper narratives. Addresses storytelling and engineering tax by enabling AI agents to process and use full research exploration.

Ax Abay Bektursun 5/20/2026

Borrowed Geometry: Cross-Distribution Head-Importance Fingerprints of Frozen Pretrained Gemma 4 31B

Studies transfer of frozen Gemma 4 31B attention head patterns across modalities. Analyzes head importance fingerprints for non-text tasks using text-pretrained weights.

Ax Junhyeong Lee, Sangjin Jin, Yongjae Lee 5/20/2026

Decision-Focused Learning via Tangent-Space Projection of Prediction Error

Decision-Focused Learning via tangent-space projection. Derives closed-form geometric characterization for regret gradients in decision-focused prediction tasks.

Ax Antonin Berthon, Nicolas Astorga, Mihaela van der Schaar 5/20/2026

Skill Neologisms: Towards Skill-based Continual Learning

Skill Neologisms approach for continual learning in LLMs. Uses soft tokens to add new skills without catastrophic forgetting or context limitations.

Ax Raimondo Fanale 5/20/2026

GRALIS: A Unified Canonical Framework for Linear Attribution Methods via Riesz Representation

GRALIS: Unified mathematical framework for XAI attribution methods (GradCAM, SHAP, LIME, Integrated Gradients). Establishes representation theory via Riesz representation theorem.

Ax Chuanyang Zheng, Jiankai Sun, Yihang Gao, Yuehao Wang, Liangchen Tan, Mac Schwager, Anderson Schneider, Yuriy Nevmyvaka, Xiaodong Liu 5/20/2026

Cubit: Token Mixer with Kernel Ridge Regression

Cubit: Token mixer alternative to attention using kernel ridge regression interpretation. Proposes replacement for transformer attention mechanism based on Nadaraya-Watson regression.

Ax Yuan Fang, Yi Xie, Xuming Ran 5/20/2026

HoReN: Normalized Hopfield Retrieval for Large-Scale Sequential Model Editing

HoReN: Codebook-based method for continual model editing in LLMs. Enables scalable factual updates to deployed models while preserving other knowledge and generalizing across paraphrases.

Ax Nikunj Gupta, James Zachary Hare, Jesse Milzman, Rajgopal Kannan, Viktor Prasanna 5/20/2026

SACHI: Structured Agent Coordination via Holistic Information Integration in Multi-Agent Reinforcement Learning

SACHI framework for multi-agent coordination in reinforcement learning. Addresses information bottleneck problem where agents must act on partial observations and local information.