Ax Nan Jia, Haojin Yang, Xing Ma, Jiesong Lian, Shuailiang Zhang, Weipeng Zhang, Ke Zeng, Xunliang Cai, Zequn Sun 5/14/2026

Asymmetric On-Policy Distillation: Bridging Exploitation and Imitation at the Token Level

Asymmetric on-policy distillation: Token-level approach addressing variance, vanishing gradients, and exploration issues in LLM training.

Ax Leonel Aguilar, Jan Nagler, Christoph Hoelscher, Nino Antulov-Fantulin 5/14/2026

Does Your Neural Network Extrapolate? Feature Engineering as Identifiability Bias for OOD Generalization

Study of when neural networks fail at OOD generalization, decoupling feature learning from data-generating-process identifiability.

Ax Joshua Shay Kricheli, Alexander Lawrence Reid, Soumajyoti Sarkar, Venkata Gandikota, Paulo Shakarian 5/14/2026

Tokens-per-Parameter Coverage Is Critical for Robust LLM Scaling Law Extrapolation

Analysis of LLM scaling laws showing tokens-per-parameter ratio affects extrapolation robustness; demonstrates collinearity causes ill-conditioning.

Ax Karim Othman, Jonas Petersen, Matei Ignuta-Ciuncanu, Camilla Mazzoleni, Federico Martelli, Alessandro Lombardi, Riccardo Maggioni, Philipp Petersen 5/14/2026

FactoryNet: A Large-Scale Dataset toward Industrial Time-Series Foundation Models

FactoryNet: First large-scale industrial time-series pretraining corpus with 51M datapoints for foundation models with cross-embodiment transfer.

Ax Musa Cim, Poovaiah Palangappa, Miro Hodak, Ravi Dwivedula, Meena Arunachalam, Mahmut Taylan Kandemir 5/14/2026

Pretraining large language models with MXFP4 on Native FP4 Hardware

Study of MXFP4 quantization for full-pipeline FP4 training of LLMs, analyzing divergence in forward/backward passes on native FP4 hardware.

Ax Daniel Goldstein, Eugene Cheah 5/14/2026

Key-Value Means: Transformers with Expandable Block-Recurrent Compressed Memory

Key-Value Means: A block-recurrent attention mechanism enabling O(N) transformers with fixed or expandable memory for long-context tasks.

Ax Debashis Guha 5/14/2026

Consolidation-Expansion Operator Mechanics:A Unified Framework for Adaptive Learning

Consolidation-Expansion Operator Mechanics: A framework for understanding adaptive learning systems through order-gap formalism.

Ax Huilin Zhou, Jian Zhao, Yilu Zhong, Zhen Liang, Xiuyuan Chen, Yuchen Yuan, Tianle Zhang, Chi Zhang, Lan Zhang, Xuelong Li 5/14/2026

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

Metis: A framework using policy optimization to automate jailbreaking of LLMs via red teaming, treating attacks as POMDP inference.

Ax Marcin Kostrzewa, Sebastian Tomczak, Roman Furman, Anna Poberezhna, Micha{\l} Furga{\l}a, Julia Farganus, Oleksii Furman, Maciej Zi\k{e}ba 5/14/2026

V4FinBench: Benchmarking Tabular Foundation Models, LLMs, and Standard Methods on Corporate Bankruptcy Prediction

V4FinBench: A benchmark with 1M+ company-year records for evaluating tabular foundation models and LLMs on corporate bankruptcy prediction.

Ax Yaxin Du, Xiyuan Yang, Zhifan Zhou, Wanxu Liu, Zixing Lei, Zimeng Chen, Fenyi Liu, Haotian Wu, Yuzhu Cai, Zexi Liu, Xinyu Zhu, WenHao Wang, Linfeng Zhang, Chen Qian, Siheng Chen 5/14/2026

DataMaster: Data-Centric Autonomous AI Research

Autonomous AI system for data-centric machine learning that searches, adapts, and validates datasets to improve model performance.

Ax Jiaming Li, Chenyu Zhu, Nanxi Yi, Youjun Bao, Li Sun, Quanying Lv, Xiang Fang, Daizong Liu, Jianjun Li, Kun He, Bowen Zhou, Zhiyuan Ma 5/14/2026

TMPO: Trajectory Matching Policy Optimization for Diverse and Efficient Diffusion Alignment

Trajectory matching policy optimization for diffusion model alignment that prevents reward hacking through probability distribution constraints.

Ax Jonas Petersen, Gian-Alessandro Lombardi, Riccardo Maggioni, Camilla Mazzoleni, Federico Martelli, Philipp Petersen 5/14/2026

HEPA: A Self-Supervised Horizon-Conditioned Event Predictive Architecture for Time Series

Self-supervised architecture for rare event prediction in multivariate time series using causal transformers and joint-embedding predictive learning.

Ax Abhishek Moturu, Anna Goldenberg, Babak Taati 5/14/2026

LiBaGS: Lightweight Boundary Gap Synthesis for Targeted Synthetic Data Selection

Lightweight method for selecting targeted synthetic training data by scoring samples on boundary proximity, uncertainty, and real-data density.

Ax Yizhu Jiao, Ruixiang Zhang, Richard Bai, Jiawei Han, Ronan Collobert, Yizhe Zhang 5/14/2026

Primal Generation, Dual Judgment: Self-Training from Test-Time Scaling

Leverages test-time scaling comparative information as self-training signal in dual judgment space for improved code generation.

Ax Jeongsol Kim, Hongeun Kim, Jian Wang, Jong Chul Ye 5/14/2026

Gradient-Free Noise Optimization for Reward Alignment in Generative Models

Proposes ZeNO, a gradient-free optimization method for reward alignment in diffusion and flow models using noise-space optimization.

Ax Han Xiao 5/14/2026

Test-Time Compute for Dense Retrieval: Agentic Program Generation with Frozen Embedding Models

Uses agentic program search to apply test-time compute to frozen embedding models, optimizing inference programs for dense retrieval.

Ax DatologyAI, :, Siddharth Joshi, Haoli Yin, Rishabh Adiga, Haakon Mongstad, Alvin Deng, Aldo Carranza, Alex Fang, Amro Abbas, Anshuman Suri, Brett Larsen, Daniel Zayas, Darren Teh, David Schwab, Diego Kiner, Fan Pan, Jack Urbanek, Jason Lee, Jason Telanoff, Josh Wills, Kaleigh Mentzer, Luke Merrick, Maximilian B\"other, Parth Doshi, Paul Burstein, Pratyush Maini, Ties Robroek, Tony Jiang, Vidhi Jain, Vineeth Dorna, Zhengping Wang, Bogdan Gaza, Ari Morcos, Matthew Leavitt 5/14/2026

20/20 Vision Language Models: A Prescription for Better VLMs through Data Curation Alone

Demonstrates data curation alone improves vision-language model performance by 11.7pp without architecture or compute changes.

Ax Yan Jiang, Ruihong Qiu, Zi Huang 5/14/2026

Block-R1: Rethinking the Role of Block Size in Multi-domain Reinforcement Learning for Diffusion Large Language Models

Investigates block size effects in reinforcement learning for diffusion large language models with semi-autoregressive generation.

Ax Sayantani Ghosh, Amit Kumar Das, Amlan Chakrabarti 5/14/2026

A New Technique for AI Explainability using Feature Association Map

Proposes FAMeX algorithm for AI explainability using graph-theoretic feature association maps.

Ax Jeffrey T. H. Wong, Cheng Zhang, Xinye Cao, Pedro Gimenes, Christos-Savvas Bouganis, George A. Constantinides, Wayne Luk, Yiren Zhao 5/14/2026

A3 : an Analytical Low-Rank Approximation Framework for Attention

Low-rank approximation framework for compressing transformer attention layers by analyzing architectural characteristics rather than individual layer outputs.

Ax Liran Ringel, Elad Tolochinsky, Yaniv Romano 5/14/2026

Learning a Continue-Thinking Token for Enhanced Test-Time Scaling

Explores learned continue-thinking tokens to extend reasoning steps and improve LLM performance through test-time compute scaling.

Ax Omer Luxembourg, Haim Permuter, Eliya Nachmani 5/14/2026

Plan for Speed: Dilated Scheduling for Masked Diffusion Language Models

Proposes Dilated Unmasking Scheduler for faster non-autoregressive text generation in masked diffusion language models by scheduling token unmasking to avoid sequential behavior.

Ax Millicent Li, Alberto Mario Ceballos Arroyo, Giordano Rogers, Naomi Saphra, Byron C. Wallace 5/14/2026

Do Activation Verbalization Methods Convey Privileged Information?

Investigates whether activation verbalization methods (using LLMs to describe internal representations) reveal privileged information or just input information.

Ax Samuel Willis, Paul Duckworth, Jack Simons, Aleksandra Kalisz, Krisztina Sinkovics, Noam Ghenassia, Shikha Surana, Henry T. Oldroyd, Alexandru I. Stere, Dragos D Margineantu, Carl Henrik Ek, Henry Moss, Erik Bodin 5/14/2026

Sample-Efficient Optimisation over the Outputs of Generative Models

O3 method for sample-efficient optimization within generative model outputs (diffusion/flow models) for task-specific criteria.

Ax Kathy Garcia, Leyla Isik 5/14/2026

Behavioral Geometric Supervision Aligns Video Foundation Models with Human Social Perception

Aligns video foundation models with human social perception using geometric supervision of behavioral features.

Ax Hamza Cherkaoui, H\'el\`ene Halconruy, Yohan Petetin 5/14/2026

When to Transfer: Adaptive Source Selection for Positive Transfer in Linear Models

Greedy algorithm for selecting which source tasks and sample sizes to transfer in multi-source transfer learning for linear models.

Ax Yuu Jinnai 5/14/2026

Re-evaluating Minimum Bayes Risk Decoding for Automatic Speech Recognition

Evaluates Minimum Bayes Risk decoding versus beam search for automatic speech recognition tasks.

Ax Jiaming Qu, Mengtian Guo, Yue Wang 5/14/2026

Why is "Chicago" Predictive of Deceptive Reviews? Using LLMs to Discover Language Phenomena from Lexical Cues

Uses LLMs to explain what linguistic features (e.g., city names) predict deceptive reviews, leveraging language models for feature interpretation.

Ax Mario Markov (INSAIT, Sofia University "St. Kliment Ohridski"), Stefan Maria Ailuro (INSAIT, Sofia University "St. Kliment Ohridski"), Luc Van Gool (INSAIT, Sofia University "St. Kliment Ohridski"), Konrad Schindler (ETH Zurich), Danda Pani Paudel (INSAIT, Sofia University "St. Kliment Ohridski") 5/14/2026

FireScope: Wildfire Risk Raster Prediction with a Chain-of-Thought Oracle

FireScope-Bench dataset and method for predicting wildfire risk using chain-of-thought reasoning with satellite imagery and climate data.

Ax Yibo Lyu, Gongwei Chen, Rui Shao, Weili Guan, Liqiang Nie 5/14/2026

PersonalAlign: Hierarchical Implicit Intent Alignment for Personalized GUI Agent with Long-Term User-Centric Records

PersonalAlign framework aligns GUI agents with implicit user intents using long-term user records as context for improved personalization.

Ax Asen Dotsinski, Panagiotis Eustratiadis 5/14/2026

Sockpuppetting: Jailbreaking LLMs by Combining Prefilling with Optimization

Sockpuppetting combines prefill attacks with optimization to jailbreak LLMs more effectively by ensembling multiple prefill variants.

Ax Fabian Fumagalli, R. Teal Witter, Christopher Musco 5/14/2026

PolySHAP: Extending KernelSHAP with Interaction-Informed Polynomial Regression

PolySHAP improves KernelSHAP algorithm for computing Shapley values in explainable AI by using polynomial regression instead of linear approximation to reduce computational cost.

Ax Shawn Im, Changdae Oh, Zhen Fang, Sharon Li 5/14/2026

How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability

Mechanistic interpretability of transformer token associations through gradient analysis revealing semantic learning mechanisms.

Ax Adam Remaki, Christel G\'erardin, Eul\`alia Farr\'e-Maduell, Martin Krallinger, Xavier Tannier 5/14/2026

SynCABEL: Synthetic Contextualized Augmentation for Biomedical Entity Linking

SynCABEL: LLM-generated synthetic training data framework for biomedical entity linking without manual annotation.

Ax Xingyuan Hua, Sheng Yue, Xinyi Li, Yizhe Zhao, Jinrui Zhang, Ju Ren 5/14/2026

Context Learning for Multi-Agent Discussion

Multi-LLM context learning method (M2CL) addressing discussion inconsistency in multi-agent collaborative problem-solving.

Ax Byeongju Woo, Zilin Wang, Byeonghyun Pak, Sangwoo Mo, Stella X. Yu 5/14/2026

Aligning Forest and Trees in Images & Long Captions for Visually Grounded Understanding

Hierarchical vision-language learning for fine-grained understanding of long captions using part-to-whole composition analysis.

Ax Tae Soo Kim, Yoonjoo Lee, Jaesang Yu, John Joon Young Chung, Juho Kim 5/14/2026

DiscoverLLM: From Executing Intents to Discovering Them

DiscoverLLM: LLM framework for discovering user intents through interactive exploration rather than clarification questions.

Ax Nikolas Melissaris, Antigoni Polychroniadou, Akira Takahashi, Chenkai Weng, Jiayi Xu 5/14/2026

ZKBoost: Zero-Knowledge Verifiable Training for XGBoost

ZKBoost: zero-knowledge proof protocol for verifying XGBoost model training integrity without revealing data or parameters.

Ax Jessie Yuan, Yilin Wu, Andrea Bajcsy 5/14/2026

When to Act, Ask, or Learn: Uncertainty-Aware Policy Steering

Uncertainty-aware policy steering for robot deployment using learned verifiers and vision-language models to select safe actions.

Ax Baoqing Yue, Zihan Zhu, Yutong Han, Qian Sun, Jichen Feng, Hufei Yang, Yifan Zhang, Mengdi Wang 5/14/2026

Interactive Benchmarks

Interactive Benchmarks: unified evaluation paradigm assessing model reasoning through adaptive information acquisition rather than fixed benchmarks.

Ax Leo Benac, Jonas Raedler, Zilin Ma, Finale Doshi-Velez 5/14/2026

A Benchmark for Multi-Party Negotiation Games from Real Negotiation Data

Benchmark and evaluation framework for multi-party sequential negotiation games with configurable generator and baseline solvers.

Ax Tom Ulanovski, Eyal Blyachman, Maya Bechler-Speicher 5/14/2026

Improving LLM Final Representations with Inter-Layer Geometry

Method improving LLM predictions by leveraging complementary signals across intermediate layers using lightweight graph neural networks.

Ax Vladimir Zaigrajew, Micha{\l} Piechota, Gaspar Sekula, Pawe{\l} Gelar, Przemys{\l}aw Biecek 5/14/2026

LINE: LLM-based Iterative Neuron Explanations for Vision Models

LINE: training-free iterative approach using LLMs to generate high-level semantic explanations of individual neurons in vision models.

Ax Mohammad Sadegh Salehi, Alex Perkins, Igor Maurell, Ashkan Dabbagh, Raymond Wong 5/14/2026

AmaraSpatial-10K: A Spatially and Semantically Aligned 3D Dataset for Spatial Computing and Embodied AI

AmaraSpatial-10K: 10K+ synthetic 3D assets optimized for embodied AI and robotics with proper metric scaling and deterministic anchoring.

Ax Teng Chen, Sheng Xu, Feixiang Guo, Xiaoyu Wang, Qingqing Gu, Hongyan Li, Luo Ji 5/14/2026

Efficient Rationale-based Retrieval: On-policy Distillation from Generative Rerankers based on JEPA

Rabtriever: efficient rationale-based document retrieval via on-policy distillation from LLM-based generative rerankers with independent encoding.

Ax Pernilla Hall, Anton Ununger, Riccardo Rubei, Alessio Bucaioni 5/14/2026

Bug-Report-Driven Fault Localization: Industrial Benchmarking and Lesson Learned at ABB Robotics

Industrial study on using AI for fault localization in software systems based on textual bug reports, tested on ABB Robotics codebase.

Ax Fuyun Wang, Yuanzhi Wang, Xu Guo, Sujia Huang, Tong Zhang, Dan Wang, Hui Yan, Xin Liu, Zhen Cui 5/14/2026

Mixture Prototype Flow Matching for Open-Set Supervised Anomaly Detection

Mixture Prototype Flow Matching framework for open-set anomaly detection using continuous transformations and multi-modal normal data modeling.

Ax Venkat Srinivasan, Vishaal Jatav, Anushka Chandrababu, Geetika Sharma 5/14/2026

Gyan: An Explainable Neuro-Symbolic Language Model

Gyan neuro-symbolic language model combining transformers with symbolic reasoning for improved interpretability and reduced hallucinations.

Ax Zixuan Xie, Xinyu Liu, Shangtong Zhang 5/14/2026

MathlibPR: Pull Request Merge-Readiness Benchmark for Formal Mathematical Libraries

MathlibPR benchmark uses LLMs to assess pull request merge-readiness for Lean formal mathematics library, addressing review bottleneck.

Ax Shota Fujikawa, Issei Sato 5/14/2026

Max-pooling Network Revisited: Analyzing the Role of Semantic Probability in Multiple Instance Learning for Hallucination Detection

Method for efficient LLM hallucination detection using multiple instance learning, improving computational efficiency of semantic consistency checks.