Ax Anne E. Staples 4/27/2026

Insect-inspired modular architectures as inductive biases for reinforcement learning

Insect-inspired distributed neural architectures as inductive biases for continuous control reinforcement learning.

Ax Emil Ryd, Henning Bartsch, Julian Stastny, Joe Benton, Vivek Hebbar 4/27/2026

Removing Sandbagging in LLMs by Training with Weak Supervision

Training method using weak supervision to prevent sandbagging where capable models underperform verification ability.

Ax Tiffany Bao, Kylie Trousil, Quang Duy Tran, Fabio Di Troia, Younghee Park 4/27/2026

Generating Synthetic Malware Samples Using Generative AI

Generative AI methods for creating synthetic malware samples with obfuscation techniques for security research.

Ax Mahdi Kallel, Johannes T\"olle, Ahmed Hendawy, Carlo D'Eramo 4/27/2026

Do Not Imitate, Reinforce: Iterative Classification via Belief Refinement

Reinforced Iterative Classification framework allowing variable compute budgets and refined predictions via iterative refinement.

Ax Harsh Kumar, Rahul Maity, Tanmay Joshi, Aman Chadha, Vinija Jain, Suranjana Trivedy, Amitava Das 4/27/2026

PermaFrost-Attack: Stealth Pretraining Seeding(SPS) for planting Logic Landmines During LLM Training

PermaFrost-Attack: poisoning attack via stealth pretraining seeding to plant backdoors in LLM training data.

Ax Meghana Karnam, Ananya Joshi 4/27/2026

Reliable Self-Harm Risk Screening via Adaptive Multi-Agent LLM Systems

Statistical framework for multi-agent LLM systems to detect unreliable decisions in self-harm risk screening.

Ax Weiqiu You, Cassandra Goldberg, Amin Madani, Daniel A. Hashimoto, Eric Wong 4/27/2026

Sum-of-Checks: Structured Reasoning for Surgical Safety with Large Vision-Language Models

Sum-of-Checks framework decomposing vision-language model reasoning for reliable surgical safety assessment.

Ax Rico Angell, Raghav Singhal, Zachary Horvitz, Zhou Yu, Rajesh Ranganath, Kathleen McKeown, He He 4/27/2026

Estimating Tail Risks in Language Model Output Distributions

Framework for estimating tail risks and rare harmful outputs in LLM distributions at population scale deployment.

Ax Zhancun Mu, Guangyu Zhao, Yiwu Zhong, Chi Zhang 4/27/2026

Preserve Support, Not Correspondence: Dynamic Routing for Offline Reinforcement Learning

Dynamic routing technique for offline reinforcement learning that balances Q-value improvement with dataset support constraints.

Ax Dharshan Kumaran, Viorica Patraucean, Simon Osindero, Petar Velickovic, Nathaniel Daw 4/27/2026

How LLMs Detect and Correct Their Own Errors: The Role of Internal Confidence Signals

Research on how LLMs detect and correct their own errors using second-order confidence signals from decision neuroscience.

Ax Marco Obermeier, Marco Pruckner, Florian Haselbeck, Andreas Zeiselmair 4/27/2026

FETS Benchmark: Foundation Models Outperform Dataset-specific Machine Learning in Energy Time Series Forecasting

FETS Benchmark: Foundation models outperform dataset-specific approaches for generalizable energy time series forecasting.

Ax Nanae Aratake, Taisei Tosaki, Yuji Okamoto, Eiichiro Uchino, Masaki Nakamura, Nobutomo Matsui, Akiko Hatakama, Yasushi Okuno 4/27/2026

A Nationwide Japanese Medical Claims Foundation Model: Balancing Model Scaling and Task-Specific Computational Efficiency

Japanese medical foundation model balancing scaling laws and task-specific efficiency for clinical risk prediction on longitudinal data.

Ax Kang Liu, Jianchen Hu 4/27/2026

SOC-ICNN: From Polyhedral to Conic Geometry for Learning Convex Surrogate Functions

SOC-ICNN: Input convex neural network architecture generalizing from linear programming to second-order cone programming.

Ax Benedikt Franke, Nils F\"orster, Frank K\"oster, Asja Fischer, Markus Lange, Arne Raulf 4/27/2026

Revisiting Neural Activation Coverage for Uncertainty Estimation

Extension of neural activation coverage technique for uncertainty estimation in regression tasks of pre-trained neural networks.

Ax Yuelin Hu, Zhenbo Yu, Zhengxue Cheng, Wei Liu, Li Song 4/27/2026

Hidden Failure Modes of Gradient Modification under Adam in Continual Learning, and Adaptive Decoupled Moment Routing as a Repair

Hidden failure modes of gradient modification under Adam optimizer in continual learning with adaptive decoupled moment routing solution.

Ax Qinhan Hou, Jing Tang 4/27/2026

Distance-Misaligned Training in Graph Transformers and Adaptive Graph-Aware Control

Analysis of distance-misaligned training failure modes in graph transformers with adaptive control mechanism.

Ax Abhinaba Basu 4/27/2026

HubRouter: A Pluggable Sub-Quadratic Routing Primitive for Hybrid Sequence Models

HubRouter: Sub-quadratic routing module replacing O(n²) attention with O(nM) hub-mediated routing for hybrid sequence models.

Ax Hojjat Karami, David Atienza, Jean-Philippe Thiran, Anisoara Ionescu 4/27/2026

FeatEHR-LLM: Leveraging Large Language Models for Feature Engineering in Electronic Health Records

FeatEHR-LLM: Framework using LLMs for automated feature engineering in electronic health records with irregular temporal patterns.

Ax Jichao Wang, Liuyang Bian, Yufeng Zhou, Han Xiao, Yue Pan, Guozhi Wang, Hao Wang, Zhaoxiong Wang, Yafei Wen, Xiaoxin Chen, Shuai Ren, Lingfang Zeng 4/27/2026

SOLAR-RL: Semi-Online Long-horizon Assignment Reinforcement Learning

SOLAR-RL: Semi-online reinforcement learning framework for training multimodal LLM-based GUI agents on complex navigation tasks.

Ax Asim Ukaye, Mubarak Abdu-Aguye, Nurbek Tastan, Karthik Nandakumar 4/27/2026

Data-Free Contribution Estimation in Federated Learning using Gradient von Neumann Entropy

Data-free contribution estimation in federated learning using gradient von Neumann entropy to identify client importance without privacy leakage.

Ax Yuqi Pan, Jinghao Zhuang, Yupeng Feng, Fangzhi Zhong, Siyu Ding, Xuerui Qiu, Shaowei Gu, Bohan Sun, Zhiyong Qin, Yibo Zhong, Lingtao Ouyang, Kun Yang, Zehao Liu, Yuhong Chou, Shurong Wang, Anjie Hu, Han Xu, Bo Xu, Guoqi Li 4/27/2026

SpikingBrain2.0: Brain-Inspired Foundation Models for Efficient Long-Context and Cross-Platform Inference

SpikingBrain2.0: 5B brain-inspired foundation model using spiking neural networks for efficient long-context inference with reduced training overhead.

Ax Bilal Faye, Abdoulaye Mbaye, Hanane Azzag, Mustapha Lebbah 4/27/2026

Adaptive Head Budgeting for Efficient Multi-Head Attention

Proposes adaptive head budgeting mechanism for multi-head attention in Transformers to improve efficiency by selectively activating attention heads based on task requirements.

Ax Jose Geraldo Fernandes, Luiz Facury, Pedro Robles Dutenhefner, Wagner Meira Jr 4/27/2026

Beyond Patient Invariance: Learning Cardiac Dynamics via Action-Conditioned JEPAs

Self-supervised learning approach using action-conditioned world models for cardiac disease detection, shifting from invariance-based to dynamics-aware objectives.

Ax In\^es Oliveira e Silva, S\'ergio Jesus, Iker Perez, Rita P. Ribeiro, Carlos Soares, Hugo Ferreira, Pedro Bizarro 4/27/2026

Rethinking XAI Evaluation: A Human-Centered Audit of Shapley Benchmarks in High-Stakes Settings

Research evaluating Shapley value variants for explainable AI in high-stakes applications, comparing theoretical formulations and human utility alignment.

Ax Yuchen Xiong, Swee Keong Yeap, Zhen Hong Ban 4/27/2026

Operational Feature Fingerprints of Graph Datasets via a White-Box Signal-Subspace Probe

WG-SRC white-box probe diagnosing graph neural network mechanisms and dataset feature requirements for node classification.

Ax Hillary Mutisya, John Mugane 4/27/2026

Neural Recovery of Historical Lexical Structure in Bantu Languages from Modern Data

Transformer model extracting historical lexical structure and cognates from modern Bantu language morphological data.

Ax Sijie Li, Shanda Li, Haowei Lin, Weiwei Sun, Ameet Talwalkar, Yiming Yang 4/27/2026

Spend Less, Fit Better: Budget-Efficient Scaling Law Fitting via Active Experiment Selection

Active experiment selection method for budget-aware scaling law fitting, reducing costs of planning large-scale ML training runs.

Ax Michael Cooper, Samuel Cooper 4/27/2026

Math Takes Two: A test for emergent mathematical reasoning in communication

Evaluation framework testing whether LLMs perform genuine mathematical reasoning vs pattern matching on novel abstract mathematical problems.

Ax Charles Junichi McAndrews 4/27/2026

Feedback Over Form: Why Execution Feedback Matters More Than Pipeline Topology in 1-3B Code Generation

Code generation pipeline study showing execution feedback outweighs topology complexity in 1-3B model composition on HumanEval.

Ax Aladin Djuhera, Haris Gacanin, Holger Boche 4/27/2026

MambaCSP: Hybrid-Attention State Space Models for Hardware-Efficient Channel State Prediction

MambaCSP: Hybrid state-space model combining attention and Mamba for hardware-efficient channel prediction with subquadratic scaling.

Ax Deepank Girish, Yi Hao Chan, Sukrit Gupta, Jing Xia, Jagath C. Rajapakse 4/27/2026

Foundation models for discovering robust biomarkers of neurological disorders from dynamic functional connectivity

RE-CONFIRM framework evaluating robustness of brain biomarkers extracted by foundation models from dynamic functional connectivity data.

Ax Zhuonan Yang, Jacob Xiaochen Li, Francisco Piedrahita Velez, Eric Todd, David Bau, Michael L. Littman, Stephen H. Bach, Ellie Pavlick 4/27/2026

Shared Lexical Task Representations Explain Behavioral Variability In LLMs

LLM prompt sensitivity investigation comparing instruction-based vs example-based prompting through shared lexical task representations.

Ax Xiaodi Li, Yang Xiao, Munhwan Lee, Konstantinos Leventakos, Young J. Juhn, David Jones, Terence T. Sio, Wei Liu, Maria Vassilaki, Nansu Zong 4/27/2026

Lightweight Retrieval-Augmented Generation and Large Language Model-Based Modeling for Scalable Patient-Trial Matching

Lightweight RAG and LLM approach for clinical patient-trial matching over heterogeneous EHR data with improved scalability.

Ax Hector Borobia, Elies Segu\'i-Mas, Guillermina Tormo-Carb\'o 4/27/2026

Where Should LoRA Go? Component-Type Placement in Hybrid Language Models

LoRA adapter placement study in hybrid language models combining attention and recurrent components (Qwen, Falcon architectures).

Ax Jun He, Deying Yu 4/27/2026

Sovereign Agentic Loops: Decoupling AI Reasoning from Execution in Real-World Systems

Sovereign Agentic Loops: Control-plane architecture decoupling LLM reasoning from execution to improve safety in API-calling agents by emitting structured intents instead of direct outputs.

Ax Tania Haghighi, Sina Gholami, Hamed Tabkhi, Minhaj Nur Alam 4/27/2026

Anatomy-Aware Unsupervised Detection and Localization of Retinal Abnormalities in Optical Coherence Tomography

ArXiv paper on unsupervised anomaly detection for retinal OCT imaging without labeled data. Medical imaging ML research.

Ax Mat\'ias Carrasco, Alejandro Cholaquidis 4/27/2026

Concave Statistical Utility Maximization Bandits via Influence-Function Gradients

ArXiv paper on multi-armed bandits optimizing statistical utility functionals via influence-function gradients. Theoretical ML research.

Ax Spencer Hutchinson, Nanfei Jiang, Mahnoosh Alizadeh 4/27/2026

Near-Optimal Regret for the Safe Learning-based Control of the Constrained Linear Quadratic Regulator

ArXiv paper on adaptive control for constrained linear quadratic regulator achieving optimal regret bounds. Control theory and learning research.

Ax Harry Dong, Timofey Efimov, Megna Shah, Jeff Simmons, Sean Donegan, Marc De Graef, Yuejie Chi 4/27/2026

Multimodal Diffusion to Mutually Enhance Polarized Light and Low Resolution EBSD Data

ArXiv paper using multimodal diffusion models to enhance polarized light and EBSD microscopy data. Domain-specific ML application.

Ax Leah Chen, Keni Chih-Hua Wu, Boon Tat Chia, Xiuqing Xing, Jian Cheng Wong 4/27/2026

AI-Driven Performance-to-Design Generation and Optimization of Marine Propellers

ArXiv paper on generative AI for marine propeller design using physics-based data generation. Applied ML research for engineering design.

Ax Xiaoyun Qiu, Yang Yu, Haifeng Xu 4/27/2026

On Benchmark Hacking in ML Contests: Modeling, Insights and Design

ArXiv paper analyzing benchmark hacking in ML contests—gaming evaluation metrics without genuine improvement. Research on evaluation methodologies.

Ax Yunho Kim, Quan Nguyen, Taewhan Kim, Youngjin Heo, Joonho Lee 4/27/2026

Learning-augmented robotic automation for real-world manufacturing

ArXiv paper on learning-augmented robotic control for manufacturing with real-world production validation. Applied ML research for industrial automation.

Ax Yifan Guo, Jann Spiess 4/27/2026

Algorithmic Feature Highlighting for Human-AI Decision-Making

ArXiv paper on algorithmic feature highlighting for human-AI decision-making systems. Research on interpretable AI agent design.

Ax Chang Sun, Zhiqiang Que, Bakhtiar Zadeh, Qibin Liu, Kevin H. Alvarez, Wayne Luk, Maria Spiropulu 4/27/2026

HGQ-LUT: Fast LUT-Aware Training and Efficient Architectures for DNN Inference

ArXiv paper on LUT-aware neural network training for FPGA inference with ultra-low latency. Hardware-efficient ML architecture research.

Ax Akram Erraqabi, Michal Valko, Alexandra Carpentier, Odalric-Ambrym Maillard 4/27/2026

Pliable rejection sampling

ArXiv paper on pliable rejection sampling with kernel estimators for sampling difficult distributions. Machine learning research with new approach.

Ax Daniele Calandriello, Alessandro Lazaric, Michal Valko 4/27/2026

Pack only the essentials: Adaptive dictionary learning for kernel ridge regression

ArXiv paper on adaptive dictionary learning for kernel ridge regression to reduce space complexity. Machine learning research addressing scalability.

Ax Zhanli Wu, Fabrizio Leisen, Miguel-Angel Luque-Fernandez, F. Javier Rubio 4/27/2026

Conformalized Super Learner

ArXiv paper on Conformalized Super Learner ensemble method with interval prediction uncertainty quantification. Machine learning research with theoretical contributions.

Ax Alberto Messina, Stefano Scotta 4/27/2026

Introducing Background Temperature to Characterise Hidden Randomness in Large Language Models

Research formalizing hidden randomness in LLMs via background temperature concept, analyzing nondeterminism from batch variation, kernel issues, and floating-point arithmetic.

Ax Xirui Li, Ming Li, Yunze Xiao, Ryan Wong, Dianqi Li, Timothy Baldwin, Tianyi Zhou 4/27/2026

Superminds Test: Actively Evaluating Collective Intelligence of Agent Society via Probing Agents

Empirical evaluation of collective intelligence in large-scale LLM agent societies using MoltBook platform with 2M+ agents, testing emergence of intelligence at scale.

Ax Marco Pegoraro, Jonas Seng, Dustin Heller, Wil M. P. van der Aalst, Kristian Kersting 4/27/2026

All Eyes on the Workflow: Automated and Efficient Event Discovery from Video Streams

SnapLog approach extracts event data from video streams using image embeddings for process mining and business process management applications.