Ax Abdelhakim Ziani (MICS), Andras Horvath (UNITO), Paolo Ballarini (MICS) 5/20/2026

Markov Chain Decoders Overcome the Heavy-Tail Limitations of Lipschitz Generative Models

Markov Chain decoders overcome Lipschitz generative model limitations for heavy-tailed distributions. Improves VAE and Lipschitz network outputs.

Ax Aritra Ghosh, Drew Oldag, Michael Tauraso, Andrew J. Connolly, Peter Ferguson, Derek Jones, Gourav Khullar, Argyro Sasli, Samarth Venkatesh, Gracia Wang, Maxine West, Dylan Berry, Neven Caplar, Colin Orion Chandler, Tanawan Chatchadanoraset, Michael W. Coughlin, Melissa DeLucchi, Alexandra Junell, Diego Miura, Felipe Fontinele Nunes, Wilson Beebe, Doug Branton, Sandro Campos, Liam Cunningham, Mi Dai, Jeremy Kubica, Konstantin Malanchev, Rachel Mandelbaum, Sean McGuire, Imad Pasha, Dan S. Taranu, Tianqing Zhang 5/20/2026

Hyrax: An Extensible Framework for Rapid ML Experimentation and Unsupervised Discovery in the Era of Rubin, Roman, and Euclid

Hyrax: open-source modular Python framework for ML lifecycle in astronomy, supporting data acquisition through deployment on GPU infrastructure.

Ax Ahmad Yehia, Abduallah Mohamed, Tianyi Wang, Jiseop Byeon, Kun Qian, Junfeng Jiao, Christian Claudel 5/20/2026

EgoTraj: Real-World Egocentric Human Trajectory Dataset for Multimodal Prediction

EgoTraj: open egocentric trajectory dataset from Meta Quest Pro for multimodal human motion prediction in robotics and navigation.

Ax Anis Radianis 5/20/2026

Learn-by-Wire Training Control Governance: Bounded Autonomous Training Under Stress for Stability and Efficiency

LBW-Guard: autonomous training control governance layer for stable LLM training under aggressive conditions. Operates above AdamW optimizer.

Ax Thiago R. Ramos, Helton Graziadei, Luben M. C. Cabezas 5/20/2026

Conformal Prediction via Transported Beta Laws

Novel conformal prediction method via transported Beta laws for calibration-conditional coverage guarantees in finite-sample settings.

Ax Ehsan Ahmadi, Hunter Schofield, Behzad Khamidehi, Fazel Arasteh, Jinjun Shan, Lili Mou, Dongfeng Bai, Kasra Rezaee 5/20/2026

RLFTSim: Realistic and Controllable Multi-Agent Traffic Simulation via Reinforcement Learning Fine-Tuning

RLFTSim: reinforcement learning framework for training realistic multi-agent traffic simulators via fine-tuning on real-world data.

Ax Lorenzo Bonin, Francesco Giacomarra, Luca Bortolussi, Jyotirmoy V. Deshmukh, Francesca Cairoli 5/20/2026

Guiding Neuro-Symbolic Scenario Generation with Spatio-Temporal Logic

Neuro-symbolic scenario generation using spatio-temporal logic for safety-critical testing in autonomous driving systems.

Ax Zhiyuan Jerry Lin, Benjamin Letham, Samuel Dooley, Maximilian Balandat, Eytan Bakshy 5/20/2026

Embedding by Elicitation: Dynamic Representations for Bayesian Optimization of System Prompts

ReElicit: Bayesian optimization framework for tuning system prompts in LLMs using aggregate feedback and dynamic text embeddings.

Ax Yanru Wu, Jianning Wang, Chongxin Gan, Yang Li 5/20/2026

Heterogeneity-Aware Dataset Scheduling for Efficient Audio Large Language Model Training

Dataset scheduling strategy for training Audio Large Language Models that manages heterogeneity to reduce conflicting gradients and improve convergence.

Ax Xingyu Li 5/20/2026

GOAL: Graph-based Objective-Aligned Diffusion Solvers for Dynamic Multi-Objective Optimization

GOAL: diffusion-based solver for dynamic multi-objective optimization using graph neural networks and conditioned generation.

Ax Ayush Agarwal, Ansh Gandhi, Jeremy A. Collins, Omar Rayyan, Aryan Sarswat, Ranjani Koushik, Masoud Moghani, Ajay Mandlekar, Animesh Garg 5/20/2026

COBALT: Crowdsourcing Robot Learning via Cloud-Based Teleoperation with Smartphones

COBALT: Cloud-based teleoperation platform for scalable robot learning via imitation using concurrent multi-user demonstrations on GPU.

Ax John T. Halloran, Noopur S. Bhatt 5/20/2026

Be Kind, Rewrite: Benign Projections via Rewriting Defend Against LLM Data Poisoning Attacks

Defense method against LLM backdoor attacks using rewriting with benign projections to mitigate data poisoning vulnerabilities.

Ax Zhengxin Zhang, Ning Wang, Sainyam Galhotra, Claire Cardie 5/20/2026

How Far Are We From True Auto-Research?

ResearchArena evaluates autonomous research agents (Claude, GPT, Kimi) on their ability to conduct full research loops including ideation, experimentation, and paper writing.

Ax Benjamin Sterling, M\'onica F. Bugallo, Tom Tirer 5/20/2026

Reducing Diffusion Model Memorization with Higher Order Langevin Dynamics

Research on reducing memorization in diffusion models using Higher-Order Langevin Dynamics to address privacy and copyright concerns.

Ax Tiejin Chen, Longchao Da, Xiaoou Liu, Hua Wei 5/20/2026

Position: Uncertainty Quantification in LLMs is Just Unsupervised Clustering

Position paper arguing mainstream uncertainty quantification in LLMs reduces to unsupervised clustering and measures internal consistency not correctness.

Ax Xiaoou Liu, Tiejin Chen, Dengjia Zhang, Yaqing Wang, Lu Cheng, Hua Wei 5/20/2026

Diagnosing Multi-step Reasoning Failures in Black-box LLMs via Stepwise Confidence Attribution

Framework for diagnosing multi-step reasoning failures in closed-source LLMs by attributing confidence scores to individual steps.

Ax Pu Zhao, Juyi Lin, Timothy Rupprecht, Arash Akbari, Chence Yang, Rahul Chowdhury, Elaheh Motamedi, Arman Akbari, Yumei He, Chen Wang, Geng Yuan, Weiwei Chen, Yanzhi Wang 5/20/2026

PhyWorld: Physics-Faithful World Model for Video Generation

Physics-faithful world model for video generation preserving physical state constraints for training embodied AI systems.

Ax Maosong Cao, Kai Chen, Haodong Duan, Yixiao Fang, Tong Gao, Ge Jiaye, Mo Li, Hongwei Liu, Junnan Liu, Yuan Liu, Chengqi Lyu, Han Lyu, Ningsheng Ma, Zerun Ma, Yu Sun, Zhiyong Wu, Linchen Xiao, Jun Xu, Haochen Ye, Zhaohui Yu, Yike Yuan, Songyang Zhang, Yufeng Zhao, Fengzhe Zhou, Peiheng Zhou, Dongsheng Zhu, Lin Zhu, Jingming Zhuo 5/20/2026

OpenCompass: A Universal Evaluation Platform for Large Language Models

Universal evaluation platform for comprehensive LLM assessment addressing limitations of static benchmarks with dynamic evaluation methods.

Ax Thomas Vincent Howe, David Wingate 5/20/2026

Language models struggle with compartmentalization

Study showing LLMs fail to share statistical strength across distinct presentations of same concepts in different languages/formats.

Ax Md Mehrab Tanjim, Jayakumar Subramanian, Xiang Chen, Branislav Kveton, Subhojyoti Mukherjee, Anlan Zhang, Sungchul Kim, Somdeb Sarkhel, Sunav Choudhury 5/20/2026

MOCHA: Multi-Objective Chebyshev Annealing for Agent Skill Optimization

Multi-objective optimization framework for LLM agent skills under platform constraints, balancing description, instruction, and context window limits.

Ax Emmy Liu, Varun Gangal, Michael Yu, Zhuofu Tao, Karan Singh, Sachin Kumar, Steven Y. Feng 5/20/2026

HalluWorld: A Controlled Benchmark for Hallucination via Reference World Models

Controlled benchmark for measuring LLM hallucination across summarization, QA, RAG, and agentic settings with reference world models.

Ax Subba Reddy Oota, Anant Khandelwal, Khushbu Pahwa, Satya Sai Srinath Namburi, Tanmoy Chakraborty, Bapi S. Raju, Manish Gupta 5/20/2026

Brain alignment of reasoning and action representations from vision-language and action models during naturalistic gameplay

Neuroscience study aligning brain activity with vision-language and action models during interactive gameplay, bridging neuroscience and ML.

Ax Daisuke Oba, Hiroki Furuta, Naoaki Okazaki 5/20/2026

Drifting Objectives for Refining Discrete Diffusion Language Models

Training method for discrete diffusion language models using drifting objectives to improve sampling-time text generation.

Ax Carlo Romeo, Andrew D. Bagdanov 5/20/2026

ARC-RL: A Reinforcement Learning Playground Inspired by ARC Raiders

Reinforcement learning environment suite for continuous-control locomotion tasks inspired by game design rather than real robotics.

Ax Guanyu Cui, Zhewei Wei, Kun He 5/20/2026

Position: The Turing-Completeness of Real-World Autoregressive Transformers Relies Heavily on Context Management

Position paper analyzing Turing-completeness of autoregressive transformers and the critical role of context management in real-world systems.

Ax Yunzhe Zhang, Hongfu Liu, Pengyu Hong 5/20/2026

Boosting Text-to-Image Diffusion Models via Core Token Attention-Based Seed Selection

Seed selection method for text-to-image diffusion models using attention dynamics on prompt core tokens to improve generation quality.

Ax Jingshan Chen, Bochen Yu, Henrik Ebel, Peter Eberhard 5/20/2026

Learning-Accelerated Optimization-based Trajectory Planning for Cooperative Aerial-Ground Handover Missions

Neural surrogate planner using LSTM networks to accelerate trajectory optimization for cooperative UAV-UGV missions.

Ax Neil Ashton, Adam Clark, Liam Heidt, Christopher Ivey, Sanjeeb Bose, Rahul Agrawal, Konrad Goc, Rishi Ranade, Corey Adams, Peter Sharpe, Sheel Nidhan, Semit Akkurt, Daniel Leibovici, Jean Kossaifi 5/20/2026

HiLiftAeroML: High-Fidelity Computational Fluid Dynamics Dataset for High-Lift Aircraft Aerodynamics

Open-source high-fidelity CFD dataset with 1800 samples for developing AI surrogate models in aircraft aerodynamics.

Ax Giovanni S. Alberti, Damiana Lazzaro, Serena Morigi, Matteo Santacesaria, Shibo Wang 5/20/2026

Diffusion Graph Posterior Sampling for Nonlinear Inverse Problems with Application to Electrical Impedance Tomography

Deep generative models using diffusion posterior sampling for inverse problems on unstructured meshes, applied to electrical impedance tomography.

Ax Lakshya A Agrawal, Donghyun Lee, Shangyin Tan, Wenjie Ma, Karim Elmaaroufi, Rohit Sandadi, Sanjit A. Seshia, Koushik Sen, Dan Klein, Ion Stoica, Joseph E. Gonzalez, Omar Khattab, Alexandros G. Dimakis, Matei Zaharia 5/20/2026

optimize_anything: A Universal API for Optimizing any Text Parameter

optimize_anything: LLM-based universal optimization system for text parameters achieving SOTA across six diverse domains with single model.

Ax Elena C Offenberg, Dirk Keller, Mariska J Vansteensel, Zachary V Freudenburg, Nick F Ramsey, Julia Berezutskaya 5/20/2026

BCI-sift: An automated feature selection toolbox for Brain Computer Interface applications

BCI-sift: Python toolbox for automated feature selection in brain-computer interface applications with diverse optimization algorithms.

Ax Gioele Molinari, Florian Felten, Soheyl Massoudi, Mark Fuge 5/20/2026

EngiAI: A Multi-Agent Framework and Benchmark Suite for LLM-Driven Engineering Design

EngiAI is a multi-agent LLM framework and benchmark for engineering design tasks combining simulation, retrieval, and manufacturing tools.

Ax Petar Radanliev, Omar Santos, Carsten Maple, Kay Atefi 5/20/2026

Operationalising Artificial Intelligence Bills of Materials (AIBOMs) for Verifiable AI Provenance and Lifecycle Assurance

Framework extending CycloneDX standard to create AI Bills of Materials (AIBOMs) for verifiable AI system provenance and supply chain transparency.

Ax Yuxuan Gao, Megan Wang, Yi Ling Yu 5/20/2026

Distribution-Free Uncertainty Quantification for Continuous AI Agent Evaluation

Machine learning paper adapting conformal prediction methods to continuous AI agent evaluation with distribution-free uncertainty quantification.

Ax Dmitry Redko (Applied AI Institute), Albert Fazlyev (AI Talent Hub, ITMO University), Konstantin Sozykin (Applied AI Institute), Maria Ivanova (YSDA, Applied AI Institute), Evgeny Burnaev (Applied AI Institute), Egor Shvetsov (Applied AI Institute) 5/20/2026

Prior Knowledge or Search? A Study of LLM Agents in Hardware-Aware Code Optimization

Research analyzing which components of LLM agent architectures contribute most to hardware-aware code optimization, studying propose-evaluate-revise loops.

Ax Florian A. D. Burnat, Brittany I. Davidson 5/20/2026

Auditing Privacy in Multi-Tenant RAG under Account Collusion

Research paper on privacy vulnerabilities in multi-tenant RAG systems when accounts collude, identifying failures in differential privacy boundaries.

Ax Zhuohan Gu, Qizheng Zhang, Omar Khattab, Samuel Madden 5/20/2026

PEEK: Context Map as an Orientation Cache for Long-Context LLM Agents

PEEK system caching context orientation knowledge to improve LLM agent performance on recurring document and code repository tasks.

Ax Parsa Esmati, Junha Hyung, Amirhossein Dadashzadeh, Jaegul Choo, Majid Mirmehdi 5/20/2026

Probability-Conserving Flow Guidance

Novel guidance method for diffusion and flow-based generative models that preserves probability distribution geometry.

Ax Nico Pelleriti, Sree Harsha Nelaturu, Zhanke Zhou, Zongze Li, Max Zimmer, Bo Han, Sebastian Pokutta 5/20/2026

What Do Evolutionary Coding Agents Evolve?

Study analyzing what mechanisms evolutionary algorithms combined with LLMs actually discover when generating and modifying code for algorithm design.

Ax Aur\'elien Pion, Emmanuel Vazquez 5/20/2026

Goal-Oriented Lower-Tail Calibration of Gaussian Processes for Bayesian Optimization

Improves Bayesian optimization by calibrating Gaussian process predictive distributions for better exploration-exploitation trade-offs.

Ax Arnulf Jentzen, Timo Kr\"oger 5/20/2026

Convergence rates for gradient descent in the training of overparameterized artificial neural networks with piecewise affine activation

Theoretical analysis of convergence rates for gradient descent training overparameterized neural networks with piecewise affine activations.

Ax Pierre Jobic, Maxime Haddouche, Benjamin Guedj 5/20/2026

Federated Learning with Nonvacuous Generalisation Bounds

Federated learning approach using randomized predictors with PAC-Bayesian generalization bounds for privacy-preserving distributed training.

Ax Xiao Li, Wei Zhang, Zhuhong Li, Qiongxiu Li, Shei PernChua, BingZe Lee, Jinghao Cui, Yifan Huang, Xiaolin Hu 5/20/2026

Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models

Faster-GCG improves efficiency of discrete token optimization attacks against aligned LLMs through better sample efficiency than prior GCG methods.

Ax Vima Gupta, Jae Hyung Ju, Kartik Sinha, Ada Gavrilovska, Anand Padmanabha Iyer 5/20/2026

Lynx: Enabling Efficient MoE Inference through Dynamic Batch-Aware Expert Selection

Lynx system enables efficient Mixture-of-Expert model inference through dynamic batch-aware expert selection to reduce memory bottlenecks.

Ax Gal Alon, Yehuda Dar 5/20/2026

How Does Overparameterization Affect Machine Unlearning of Deep Neural Networks?

Investigates how model overparameterization affects machine unlearning performance in deep neural networks through validation-based tuning.

Ax Samuel Hurault, Matthieu Terris, Thomas Moreau, Gabriel Peyr\'e 5/20/2026

From Score Matching to Diffusion: A Fine-Grained Error Analysis in the Gaussian Setting

Theoretical analysis of error propagation in score matching and diffusion-based sampling from unknown distributions in generative modeling.

Ax Ruoyu Chen, Siyuan Liang, Jingzhi Li, Shiming Liu, Li Liu, Hua Zhang, Xiaochun Cao 5/20/2026

Less is More: Efficient Black-box Attribution via Minimal Interpretable Subset Selection

Attribution method for identifying influential input regions in black-box models through minimal interpretable subset selection with combinatorial optimization.

Ax Yoshitaka Inoue, Tianfan Fu, Augustin Luna 5/20/2026

GraphPINE: Graph Importance Propagation for Interpretable Drug Response Prediction

Graph neural network architecture for interpretable drug response prediction using domain-specific prior knowledge and importance propagation.

Ax Oliver Mortensen, Mohammad Sadegh Talebi 5/20/2026

Recursive Entropic Risk Optimization in Discounted MDPs: Sample Complexity Bounds with a Generative Model

Studies sample complexity of risk-sensitive reinforcement learning in finite MDPs using recursive entropic risk measures with a generative model.

Ax Thomas Delliaux, Nguyen-Khanh Vu, Vincent Fran\c{c}ois-Lavet, Elise van der Pol, Emmanuel Rachelson 5/20/2026

Learning Abstract World Models with a Group-Structured Latent Space

Proposes learning abstract world models for MDPs using group-structured latent spaces with geometric priors to improve generalization from limited data.