Ax Tianyu Li, Dongchen Han, Zixuan Cao, Haofeng Huang, Mengyu Zhou, Ming Chen, Erchao Zhao, Xiaoxi Jiang, Guanjun Jiang, Gao Huang 5/22/2026

SiameseNorm: Breaking the Barrier to Reconciling Pre/Post-Norm

SiameseNorm architecture reconciles pre-norm and post-norm trade-offs in Transformers for improved stability and capacity.

Ax Beicheng Xu, Keyao Ding, Wei Liu, Yupeng Lu, Bin Cui 5/22/2026

CoFEH: LLM-driven Feature Engineering Empowered by Collaborative Bayesian Hyperparameter Optimization

CoFEH: LLM-driven automated feature engineering with Bayesian hyperparameter optimization for flexible ML pipelines.

Ax Rosie Zhao, Anshul Shah, Xiaoyu Zhu, Xinke Deng, Zhongyu Jiang, Yang Yang, Joerg Liebelt, Arnab Mondal 5/22/2026

On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs

Analysis of robustness and consistency issues in RL-finetuned vision-language models under visual perturbations and hallucinations.

Ax Filippo Rinaldi, Aniello Panariello, Giacomo Salici, Angelo Porrello, Simone Calderara 5/22/2026

Transporting Task Vectors across Different Architectures without Training

Theseus: training-free method to transfer task-specific parameter updates across models with different architectures and widths.

Ax Edward Chen, Sanmi Koyejo, Carlos Guestrin 5/22/2026

Discovering Implicit Large Language Model Alignment Objectives

Method to discover and interpret implicit alignment objectives in LLMs without pre-defined rubrics, addressing reward hacking risks.

Ax Yunshu Wu, Jiayi Cheng, Longxuan Yu, Partha Thakuria, Rob Brekelmans, Evangelos E. Papalexakis, Greg Ver Steeg 5/22/2026

Discrete Stochastic Localization for Non-autoregressive Generation

Discrete Stochastic Localization framework for non-autoregressive sequence generation using continuous-state diffusion.

Ax Ziqiao Shang, Lingyue Ge, Zi-Jian Cheng, Shi-Yu Tian, Zhenyu Huang, Wenbo Fu, Weiming Wu, Yang Chen, Xiangwen Zhang, Yulan Hu, Bin Liu, Yu-Feng Li, Lan-Zhe Guo 5/22/2026

MapTab: Are MLLMs Ready for Multi-Criteria Route Planning in Heterogeneous Graphs?

MapTab benchmark evaluates multimodal LLMs on multi-criteria route planning reasoning tasks in heterogeneous graphs.

Ax Vamshi Sunku Mohan, Kaustubh Gupta, Aneesha Das, Chandan Singh 5/22/2026

Interpreting and Steering State-Space Models via Activation Subspace Bottlenecks

arXiv paper on interpreting and steering state-space models (Mamba) using activation subspace bottlenecks and mechanistic interpretability.

Ax Sayed Mohammadreza Tayaranian Hosseini, Amir Ardakani, Warren J. Gross 5/22/2026

InnerQ: Hardware-Aware Tuning-Free Quantization of KV Cache for Large Language Models

arXiv paper on InnerQ: hardware-aware quantization of KV cache for efficient LLM text generation with reduced memory footprint.

Ax Zhixia Zhang, Zixuan Huang, Gongxun Li, Huaiyang Wang, Chengyi Yuan, Xin Xia, Deqing Wang, Fuzhen Zhuang, Shuai Ma, Ning Ding, Yaodong Yang, Jianxin Li, Yikun Ban 5/22/2026

Heterogeneous Agent Collaborative Reinforcement Learning

arXiv paper on Heterogeneous Agent Collaborative RL where agents share verified rollouts during training for mutual improvement.

Ax Fengzhi Li, Liang Zhang, Yuan Zuo, Ruiqing Zhao, YanSong Liu, Yunfei Ma, Fanyu Meng, Junlan Feng 5/22/2026

Beyond One-Size-Fits-All: Adaptive Subgraph Denoising for Zero-Shot Graph Learning with Large Language Models

arXiv paper on adaptive subgraph denoising for zero-shot graph learning using LLMs as predictors, improving cross-modal alignment.

Ax Peter Adema, Karim Galliamov, Aleksey Evstratovskiy, Ross Geurts 5/22/2026

[Re] FairDICE: A Fair Tradeoff in Multi-objective Offline RL

arXiv reproduction paper of FairDICE algorithm for fair multi-objective offline reinforcement learning from demonstrations.

Ax Chen-Hao Chao, Wei-Fang Sun, Junwei Quan, Chun-Yi Lee, Rahul G. Krishnan 5/22/2026

MDM-Prime-v2: Binary Encoding and Index Shuffling Enable Scaling of Diffusion Language Models

arXiv paper on MDM-Prime-v2, scaling masked diffusion language models via binary encoding and index shuffling for improved generalization.

Ax Tianyang Luo, Tao Feng, Zhigang Hua, Yan Xie, Shuang Yang, Ge Liu, Jiaxuan You 5/22/2026

MemReward: Graph-Based Experience Memory for LLM Reward Prediction with Limited Labels

arXiv paper on MemReward: graph-based experience memory for LLM reward prediction with limited labels in reinforcement learning scenarios.

Ax Guillaume Baudart, Marc Lelarge, Tristan St\'erin, Jules Viennot 5/22/2026

Putnam 2025 Problems in Rocq using Opus 4.6 and Rocq-MCP

Claude Opus 4.6 with Rocq proof assistant tools via Model Context Protocol autonomously solved 10/12 Putnam 2025 competition problems.

Ax Abdou-Raouf Atarmla 5/22/2026

Rule-State Inference (RSI): A Bayesian Framework for Compliance Monitoring in Rule-Governed Domains

arXiv paper on Bayesian framework for compliance monitoring in rule-governed domains without labeled outcomes, handling missing observations.

Ax Dongxia Wu, Yuhui Zhang, Serena Yeung-Levy, Emma Lundberg, Emily B. Fox 5/22/2026

Uncertainty-Aware Distribution-to-Distribution Flow Matching for Scientific Imaging

arXiv paper on uncertainty-aware generative models for scientific imaging using stochastic flow matching with out-of-distribution detection.

Ax Dongxia Wu, Shiye Su, Yuhui Zhang, Elaine Sui, Emma Lundberg, Emily B. Fox, Serena Yeung-Levy 5/22/2026

CellFluxRL: Biologically-Constrained Virtual Cell Modeling via Reinforcement Learning

arXiv paper using reinforcement learning to train biological constraint-respecting generative models for virtual cell simulation.

Ax Renato Cordeiro de Amorim, Vladimir Makarenkov 5/22/2026

Cluster-Adaptive Feature Extraction and its Theoretical Foundation with Minkowski Weighted k-Means

arXiv paper extending k-means clustering with Minkowski distance and feature weighting for adaptive feature extraction.

Ax Alexander Brenning, Thomas Suesse 5/22/2026

Aligning Validation with Deployment in Spatial Prediction: Target-Weighted Cross-Validation

arXiv paper on target-weighted cross-validation for spatial prediction models addressing biased performance estimation.

Ax Pavel Golikov, Evgenii Opryshko, Gennady Pekhimenko, Mark C. Jeffrey 5/22/2026

Robust Reasoning Benchmark

arXiv paper introducing Robust Reasoning Benchmark testing 8 LLMs against textual perturbations on AIME math problems.

Ax Amrut Nadgir, Vijay Balasubramanian, Pratik Chaudhari 5/22/2026

How does Chain of Thought decompose complex tasks?

Analysis of how chain-of-thought reasoning decomposes complex tasks by breaking classification into smaller subtasks to reduce error.

Ax Jiabao Brad Wang, Xiang Shi, Yiliang Yuan, Mustafa Misir 5/22/2026

GeoPAS: Geometric Probing for Algorithm Selection in Continuous Black-Box Optimization

Geometric probing framework for automated algorithm selection in continuous black-box optimization using landscape sampling.

Ax Azza Fadhel, The Hung Tran, Trong Nghia Hoang, Jana Doppa 5/22/2026

Black-Box Optimization From Small Offline Datasets via Meta Learning with Synthetic Tasks

Meta-learning approach for offline black-box optimization from small datasets, targeting molecular and materials discovery.

Ax Jing Sun 5/22/2026

Representation over Routing: Overcoming Surrogate Hacking in Multi-Timescale PPO

Study on multi-timescale credit assignment in PPO reinforcement learning, addressing surrogate hacking through representation-based approaches.

Ax Yuanda Xu, Hejian Sang, Zhengze Zhou, Ran He, Zhipeng Wang, Alborz Geramifard 5/22/2026

TIP: Token Importance in On-Policy Distillation

Research on token importance in on-policy knowledge distillation for LLMs, identifying which token positions provide useful learning signals during training.

Ax Stefano Colamonaco, David Debot, Pietro Barbiero, Giuseppe Marra 5/22/2026

Prototype-Grounded Concept Models for Verifiable Concept Alignment

Proposes Prototype-Grounded Concept Models grounding interpretable concepts in learned visual prototypes for verifiable concept alignment in deep learning.

Ax Xinrun Wang, Deshun Xia, Yuxi Sun, Weijie Zhu 5/22/2026

SceneSelect: Selective Learning for Trajectory Scene Classification and Expert Scheduling

Introduces SceneSelect for selective trajectory prediction across heterogeneous scenes via expert scheduling and scene classification.

Ax Yongquan Yang 5/22/2026

Negative Ontology of True Target for Machine Learning: Towards Evaluation and Learning under Democratic Supervision

Philosophical examination of true target existence assumptions in ML paradigms proposing framework for evaluation under democratic supervision.

Ax Bernd Frauenknecht, Lukas Kesper, Daniel Mayfrank, Henrik Hose, Sebastian Trimpe 5/22/2026

Uncertainty-Aware Predictive Safety Filters for Probabilistic Neural Network Dynamics

Proposes uncertainty-aware predictive safety filters using probabilistic neural networks for enforcing constraints during deep reinforcement learning.

Ax Mohammad Hossein Moslemi, Nima Hosseini Dashtbayaz, Zhimin Mei, Bissan Ghaddar, Boyu Wang 5/22/2026

Fair Dataset Distillation via Cross-Group Barycenter Alignment

Addresses fairness in dataset distillation via cross-group barycenter alignment to preserve predictive signals for demographic subgroups.

Ax Sai Niranjan Ramachandran, Suvrit Sra 5/22/2026

Trees to Flows and Back: Unifying Decision Trees and Diffusion Models

Establishes mathematical correspondence between decision trees and diffusion models via Global Trajectory Score Matching optimization principle.

Ax Sofianos Panagiotis Fotias, Vassilis Gaganis 5/22/2026

Inducing Permutation Invariant Priors in Bayesian Optimization for Carbon Capture and Storage Applications

Applies permutation invariant Bayesian optimization to carbon capture problems, improving Gaussian Process surrogate models for unordered set inputs.

Ax Shu-Hao Zhang, Le-Tong Huang, Xiang-Sheng Deng, Xin-Yi Zou, Chen Wu, Nan Li, Shao-Qun Zhang, Zhi-Hua Zhou 5/22/2026

EdgeRazor: A Lightweight Framework for Large Language Models via Mixed-Precision Quantization-Aware Distillation

Presents EdgeRazor framework compressing LLMs via mixed-precision quantization-aware distillation for sub-4-bit deployment on resource-constrained devices.

Ax Yaobo Zhang 5/22/2026

Jordan-RoPE: Non-Semisimple Relative Positional Encoding via Complex Jordan Blocks

Proposes Jordan-RoPE extending relative positional encodings using complex Jordan blocks and nilpotent operators for improved transformer attention mechanisms.

Ax Ioannis Chatzis, Jason Liartis, Athanasios Voulodimos, Giorgos Stamou 5/22/2026

Optimal Recourse Summaries via Bi-Objective Decision Tree Learning

Proposes bi-objective decision tree learning for generating actionable recourse summaries enabling global auditing and bias detection across population subgroups.

Ax Yaniv Oren, Viliam Vadocz, Joery A. de Vries, Wendelin B\"ohmer, Matthijs T. J. Spaan, Hendrik Baier 5/22/2026

PMCTS: Particle Monte Carlo Tree Search for Principled Parallelized Inference Time Scaling

Introduces Particle MCTS enabling principled parallel Monte Carlo tree search for neural network inference scaling with preserved sample efficiency.

Ax Zhiming Huang, Jamie Morgenstern, Aaron Roth, Claire Jie Zhang 5/22/2026

Instance-Adaptive Online Multicalibration

Develops online multicalibration algorithm adapting between benign and worst-case sequences via dyadic grid refinement with optimal convergence rates.

Ax Huilin Zhou, Jian Zhao, Yilu Zhong, Zhen Liang, Xiuyuan Chen, Yuchen Yuan, Tianle Zhang, Chi Zhang, Lan Zhang, Xuelong Li 5/22/2026

Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization

Proposes Metis framework reformulating LLM jailbreaking as inference-time policy optimization using metacognitive POMDP approach for automated red teaming.

Ax Yonghan Yang, Ye Yuan, Zipeng Sun, Linfeng Du, Bowei He, Haolun Wu, Can Chen, Xue Liu 5/22/2026

Support-Proximity Augmented Diffusion Estimation for Offline Black-Box Optimization

Addresses offline black-box optimization using diffusion estimation and proximity constraints to handle out-of-distribution extrapolation in design discovery.

Ax Yuxiang Chen, Dingli Liang, Yihang Chen, Ziqin Gong, Chenyang Le, Zhaokai Wang, Jiachen Zhu, Lingyu Yang, Jianghao Lin, Weinan Zhang, Jun Wang 5/22/2026

Holder Policy Optimisation

Introduces Holder Policy Optimisation improving on GRPO by adaptively aggregating token-level probabilities for trajectory-level advantages in LLM training.

Ax Yunshu Wu, Jiayi Cheng, Longxuan Yu, Partha Thakuria, Rob Brekelmans, Evangelos E. Papalexakis, Greg Ver Steeg 5/22/2026

Discrete Stochastic Localization for Non-autoregressive Generation

Proposes Discrete Stochastic Localization for non-autoregressive generation using continuous diffusion with unit-sphere embeddings, improving over masked discrete diffusion models.

Ax Krish Sharma, Omar Naim, Soumadeep Saha, Vinija Jain, Aman Chadha, Nicholas Asher 5/22/2026

TAPIOCA: Why Task- Aware Pruning Improves OOD model Capability

Investigates task-aware layer pruning effects on model generalization. Finds pruning improves out-of-distribution accuracy while harming in-distribution performance across polynomial tasks and LLMs.

Ax John T. Robertson, Jianing Zhu, Haris Vikalo, Zhangyang Wang 5/22/2026

When Is Rank-1 Steering Cheap? Geometry, Granularity, and Budgeted Search

Analyzes rank-1 activation steering for controlling LLMs. Shows effectiveness variability reflects search difficulty rather than concept limits, formalizing steering as budget-constrained optimization.

Ax Youngin Kim, Ray Sun, Inho Kim, Bumsoo Park, Hyun Oh Song 5/22/2026

Identifiable Token Correspondence for World Models

Introduces Identifiable Token Correspondence to improve temporal consistency in token-based transformer world models for long-horizon visual RL.

Ax Arne Nix, Robert James, Lasse Borgholt, Anna B. Ekner, Lana Krumm, Julius Severin, Dan Engel, Lars Maal{\o}e, Jakob Havtorn 5/22/2026

Symphony for Speech-to-Text: Supporting Real-Time Medical Voice Interfaces

Develops speech-to-text system optimized for medical terminology, contextual ambiguity, and specialized clinical notation in healthcare applications.

Ax Egor Shvetsov, Aleksandr Serkov, Shokorov Viacheslav, Redko Dmitry, Vladislav Goloshchapov, Evgeny Burnaev 5/22/2026

Bug or Feature$^2$: Weight Drift, Activation Sparsity and Spikes

Analyzes weight drift and activation sparsity in neural network training dynamics caused by interactions between standard losses and activation functions.

Ax Muhammad Umer, Muhammad Ahmed Mohsin, Ahsan Bilal, Arslan Chaudhry, Andreas Haupt, Sanmi Koyejo, Emily Fox, John M. Cioffi 5/22/2026

General Preference Reinforcement Learning

Proposes General Preference Reinforcement Learning to unify online RL for verifiable tasks with preference optimization for open-ended generation in LLM alignment.

Ax Mridul Gupta, Samyak Jain, Vansh Ramani, Hariprasad Kodamana, Sayan Ranu 5/22/2026

Position: Graph Condensation Needs a Reset -- Move Beyond Full-dataset Training and Model-Dependence

Position paper critiquing graph condensation methods and proposing solutions beyond full-dataset training.

Ax Bariscan Bozkurt, Efe Ali Gorguner, Francesco Innocenti, Rafal Bogacz 5/22/2026

Normative Networks for Source Separation via Local Plasticity and Dendritic Computation

Biologically plausible neural networks for blind source separation using local plasticity and dendritic computation.