Ax Eichi Uehara 5/4/2026

SHIFT: Robust Double Machine Learning for Average Dose-Response Functions under Heavy-Tailed Contamination

SHIFT: Robust double machine learning estimator for dose-response functions handling heavy-tailed contamination with Welsch loss.

Ax Jugal Gajjar, Kamalasankari Subramaniakuppusamy 5/4/2026

RSAT: Structured Attribution Makes Small Language Models Faithful Table Reasoners

RSAT: Method training small LMs (1-8B) to produce step-by-step table reasoning with cell-level citations using structured JSON and GRPO.

Ax Carles Domingo-Enrich, Yuanqi Du, Michael S. Albergo 5/4/2026

A unified perspective on fine-tuning and sampling with diffusion and flow models

Unified perspective on fine-tuning and sampling diffusion and flow models via stochastic optimal control and thermodynamics.

Ax Yuhui Lu, Wenjing Liu, Kun Zhan 5/4/2026

Information-geometric adaptive sampling for graph diffusion

Information-geometric framework for adaptive sampling in graph diffusion models using Riemannian manifold geometry.

Ax Bhagyashree Wagh, Akash Singh 5/4/2026

Lost in State Space: Probing Frozen Mamba Representations

Analysis of Mamba's state space representations for extracting sentence embeddings without fine-tuning or pooling heads.

Ax Hamidreza Saghir 5/4/2026

How Language Models Process Out-of-Distribution Inputs: A Two-Pathway Framework

Research on how LLMs detect out-of-distribution inputs, revealing confounds in OOD detection methods like CED and RAUQ.

Ax Zawad Yalmie Sazid, Robert Abbas, Sasa Maric 5/4/2026

A Comparative Analysis of Machine Learning Models for Intrusion Detection in Intelligent Transport Systems

ML models for intrusion detection in intelligent transport systems using edge computing and AI for security.

Ax Rebecca Saul, Jingzhi Jiang, Elliott Chia, David Wagner 5/4/2026

Trident: Improving Malware Detection with LLMs and Behavioral Features

Uses LLMs with reasoning to process sandbox behavior reports for improved PE malware detection beyond static features.

Ax Yuxuan Gao, Megan Wang, Yi Ling Yu 5/4/2026

Token Arena: A Continuous Benchmark Unifying Energy and Cognition in AI Inference

Continuous benchmark measuring AI inference at endpoint granularity across energy, speed, and latency dimensions.

Ax Minghui Chen, Chenxu Yang, Hengjie Zhu, Dayan Wu, Zheng Lin, Qingyi Si 5/4/2026

Online Self-Calibration Against Hallucination in Vision-Language Models

Method for self-calibrating vision-language models online to reduce hallucinations without relying on external model supervision.

Ax Jieming Di, Xiaoyu Chen, Ying She, Siyu Wang, Lizzie Liu, Fenggang Wu, Jiaoying Mu, Tony Tsui, Amr Elroumy, Hsing Tang, Zewei Jiang, Qiao Yang, Lin Qi, Haibo Lin, Weifeng Cui, Daniel Li, Kapil Gupta, Shivendra Pratap Singh, Jie Zheng, Arnold Overwijk, Ling Leng, Sri Reddy, Robert Malkin, Rocky Liu 5/4/2026

Intelligent Elastic Feature Fading: Enabling Model Retrain-Free Feature Efficiency Rollouts at Scale

Production infrastructure system enabling feature efficiency rollouts without model retraining in large-scale ranking systems.

Ax Akira Tamamori 5/4/2026

Geometric analysis of attractor boundaries and storage capacity limits in kernel Hopfield networks

Analyzes attractor basin geometry and storage capacity limits in kernel Hopfield networks for associative memory.

Ax Ben Wan, Yan Feng, Zihan Tang, Weizhe Huang, Yuting Zeng, Jia Wang, Tongxuan Liu 5/4/2026

RTPrune: Reading-Twice Inspired Token Pruning for Efficient DeepSeek-OCR Inference

Token pruning method for DeepSeek-OCR vision-language model to reduce inference costs while preserving text fidelity.

Ax Bo Tang, Weiwei Xie 5/4/2026

Scalable Learning in Structured Recurrent Spiking Neural Networks without Backpropagation

Proposes structured recurrent spiking neural networks with sparse connectivity and scalable learning without backpropagation.

Ax Maksym Nechepurenko, Pavel Shuvalov 5/4/2026

Foresight Arena: An On-Chain Benchmark for Evaluating AI Forecasting Agents

On-chain benchmark for evaluating AI forecasting agents with permissionless evaluation resistant to overfitting and data contamination.

Ax Pankaj Gupta, Kartik Bose 5/4/2026

RadLite: Multi-Task LoRA Fine-Tuning of Small Language Models for CPU-Deployable Radiology AI

Fine-tunes small 3-4B parameter LLMs with LoRA for radiology tasks, enabling CPU deployment in resource-constrained environments.

Ax Weifei Jin, Xilong Wang, Wei Zou, Jinyuan Jia, Neil Gong 5/4/2026

CleanBase: Detecting Malicious Documents in RAG Knowledge Databases

CleanBase system detects malicious documents in RAG knowledge databases to prevent prompt injection attacks that compromise LLM integrity.

Ax Wenda Chu, Bingliang Zhang, Jiaqi Han, Yizhuo Li, Linjie Yang, Yisong Yue, Qiushan Guo 5/4/2026

End-to-End Autoregressive Image Generation with 1D Semantic Tokenizer

End-to-end autoregressive image generation pipeline jointly optimizing visual tokenizer and generative model using foundation models.

Ax Dongxin Guo, Jikun Wu, Siu Ming Yiu 5/4/2026

SAGA: Workflow-Atomic Scheduling for AI Agent Inference on GPU Clusters

SAGA proposes workflow-atomic scheduling for AI agent inference on GPU clusters, treating entire agent workflows as first-class objects to reduce latency.

Ax M. Grailoo, J. N\'u\~nez-Y\'a\~nez 5/4/2026

Tempus: A Temporally Scalable Resource-Invariant GEMM Streaming Framework for Versal AI Edge

Tempus framework for efficient GEMM acceleration on AMD Versal edge hardware targeting LLM inference deployment constraints.

Ax Aharon Azulay, Jan Dubi\'nski, Zhuoyun Li, Atharv Mittal, Yossi Gandelsman 5/4/2026

Jailbreaking Vision-Language Models Through the Visual Modality

Four jailbreak attacks exploiting vision modality of vision-language models to bypass safety alignment through visual encoding and substitution.

Ax Massimo Rondelli, Francesco Pivi, Maurizio Gabbrielli 5/4/2026

BlenderRAG: High-Fidelity 3D Object Generation via Retrieval-Augmented Code Synthesis

BlenderRAG system uses retrieval-augmented generation to improve LLM code synthesis for 3D object generation in Blender with reduced errors.

Ax Zihao Ding, Beining Wu, Jun Huang 5/4/2026

EASE: Federated Multimodal Unlearning via Entanglement-Aware Anchor Closure

EASE framework for federated multimodal unlearning that disentangles forgotten knowledge across modalities and decentralized clients.

Ax Theodore Papamarkou, Pierre Alquier, Matthias Bauer, Wray Buntine, Andrew Davison, Gintare Karolina Dziugaite, Maurizio Filippone, Andrew Y. K. Foong, Vincent Fortuin, Dimitris Fouskakis, Jes Frellsen, Eyke H\"ullermeier, Theofanis Karaletsos, Mohammad Emtiyaz Khan, Nikita Kotelevskii, Salem Lahlou, Yingzhen Li, Fang Liu, Clare Lyle, Thomas M\"ollenhoff, Konstantina Palla, Maxim Panov, Yusuf Sale, Kajetan Schweighofer, Artem Shelmanov, Siddharth Swaroop, Martin Trapp, Willem Waegeman, Andrew Gordon Wilson, Alexey Zaytsev 5/4/2026

Position: agentic AI orchestration should be Bayes-consistent

Position paper arguing that agentic AI system orchestration layers should use Bayesian approaches for decision-making under uncertainty with tools and experts.

Ax Indraneil Paul, Glava\v{s} Glavas, Iryna Gurevych 5/4/2026

Themis: Training Robust Multilingual Code Reward Models for Flexible Multi-Criteria Scoring

Themis introduces reward models for multilingual code generation that extend beyond execution feedback to enable multi-criteria scoring and policy alignment during LLM post-training.

Ax Xihao Chen, Yangyang Guo, Roger Zimmermann 5/4/2026

Make Your LVLM KV Cache More Lightweight

LightKV method reducing KV cache memory overhead in large vision-language models by exploiting token redundancy.

Ax Jinpai Zhao, Nishant Panda, Yen Ting Lin, Eirik Valseth, Diane Oyen, Clint Dawson 5/4/2026

HyCOP: Hybrid Composition Operators for Interpretable Learning of PDEs

Framework learning PDE solution operators by composing modular components conditioned on system regime features.

Ax Simeon Allmendinger, Domenique Zipperling, Lukas Struppek, Niklas K\"uhl 5/4/2026

CollaFuse: Collaborative Diffusion Models

Federated learning approach for training diffusion models collaboratively while addressing data privacy and computational constraints.

Ax Caitlin Ho, Andrea Arnold 5/4/2026

Dynamics-Encoded Deep Learning for Robust System Identification and Parameter Estimation

Deep learning methods combining neural networks with numerical differential equation solvers for dynamics discovery and parameter estimation.

Ax Yuxin Tian, Mouxing Yang, Yuhao Zhou, Jian Wang, Qing Ye, Tongliang Liu, Gang Niu, Jiancheng Lv 5/4/2026

Learning Locally, Revising Globally: Global Reviser for Federated Learning with Noisy Labels

Federated learning approach for handling noisy labels across heterogeneous clients with different label noise types and data distributions.

Ax Ziyuan Zhang, Darcy Wang, Ningyuan Chen, Rodrigo Mansur, Vahid Sarhangian 5/4/2026

Comparing Exploration-Exploitation Strategies of LLMs and Humans: Insights from Standard Multi-armed Bandit Experiments

Empirical study comparing exploration-exploitation tradeoff behavior of LLMs versus humans in multi-armed bandit experiments.

Ax James E. Warner, Tristan A. Shah, Patrick E. Leser, Geoffrey F. Bomarito, Joshua D. Pribe, Michael C. Stanley 5/4/2026

Latent Generative Modeling of Random Fields from Limited Training Data

Generative modeling approach for random fields from limited training data using deep generative models.

Ax Eli Chien, Wei-Ning Chen, Pan Li 5/4/2026

Privacy Amplification in Differentially Private Zeroth-Order Optimization with Hidden States

Privacy amplification analysis for zeroth-order optimization in differentially private LLM fine-tuning with memory constraints.

Ax Kundan Krishna, Joseph Y Cheng, Charles Maalouf, Leon A Gatys 5/4/2026

Disentangled Safety Adapters Enable Efficient Guardrails and Flexible Inference-Time Alignment

Disentangled Safety Adapters framework for efficient AI safety through lightweight adapters decoupled from task models.

Ax Saeed Mohseni-Sehdeh, Walid Saad, Kei Sakaguchi, Tao Yu 5/4/2026

Diffusion Models for Solving Inverse Problems via Posterior Sampling with Piecewise Guidance

Diffusion model framework for solving inverse problems using piecewise guidance from conditional distributions.

Ax Zhaomin Wu, Mingzhe Du, See-Kiong Ng, Bingsheng He 5/4/2026

Beyond Prompt-Induced Lies: Investigating LLM Deception on Benign Prompts

Study of spontaneous deception in LLMs on benign prompts without explicit hidden objectives, examining trustworthiness risks.

Ax Ming-I Huang, Chih-Duo Hong, Fang Yu 5/4/2026

Concolic Testing on Individual Fairness of Neural Network Models

PyFair framework for verifying individual fairness in DNNs using concolic testing with formal guarantees.

Ax Xi He 5/4/2026

Optimal hypersurface decision trees

Optimal decision tree algorithms addressing expressivity and scalability through non-axis-parallel splits.

Ax Zihan Zhou, Chenguang Wang, Hongyi Ye, Yongtao Guan, Tianshu Yu 5/4/2026

Incomplete Data, Complete Dynamics: A Diffusion Approach

Diffusion-based framework for learning physical dynamics from incomplete and irregularly sampled observational data.

Ax Joshua Mitton, Prarthana Bhattacharyya, Ralph Abboud, Simon Woodhead 5/4/2026

Knowing When to Defer: Selective Prediction for Responsible Knowledge Tracing

Selective prediction layer using MC-Dropout for knowledge tracing models to defer uncertain predictions to humans.

Ax Zhaomin Wu, Haodong Zhao, Ziyang Wang, Jizhou Guo, Qian Wang, Bingsheng He 5/4/2026

LLM DNA: Tracing Model Evolution via Functional Representations

Method for tracking evolutionary relationships between LLMs via functional DNA-inspired representations without tokenizer assumptions.

Ax Siva Viknesh, Amirhossein Arzani 5/4/2026

Differentiable Autoencoding Neural Operator for Interpretable and Integrable Latent Space Modeling

DIANO framework for interpretable latent space modeling in scientific machine learning using neural operators.

Ax Ali Azizpour, Madeline Navarro, Santiago Segarra 5/4/2026

Adaptive Node Feature Selection For Graph Neural Networks

Adaptive feature selection approach for graph neural networks that identifies and removes unnecessary features during training to improve interpretability.

Ax Xinshuai Dong, Ignavier Ng, Haoyue Dai, Jiaqi Sun, Xiangchen Song, Peter Spirtes, Kun Zhang 5/4/2026

Score-based Greedy Search for Structure Identification of Partially Observed Linear Causal Models

Score-based greedy search method for identifying causal structure in partially observed linear systems, addressing challenges in causal discovery.

Ax Murtaza Rangwala, Farag Azzedin, Richard O. Sinnott, Rajkumar Buyya 5/4/2026

SketchGuard: Scaling Byzantine-Robust Decentralized Federated Learning via Sketch-Based Screening

Byzantine-robust decentralized federated learning using sketch-based screening to reduce communication costs in privacy-preserving collaborative training.

Ax Lorenzo Nikiforos, Luciano Prono, Charalampos Antoniadis, Fabio Pareschi, Riccardo Rovatti, Gianluca Setti 5/4/2026

Vanishing Contributions: A Unified Framework for Smooth and Iterative Model Compression

Unified framework for iterative neural network compression combining pruning, quantization, and low-rank decomposition while mitigating accuracy degradation.

Ax Qingyao Ai, Yichen Tang, Changyue Wang, Jianming Long, Weihang Su, Yiqun Liu 5/4/2026

MemoryBench: A Benchmark for Memory and Continual Learning in LLM Systems

MemoryBench benchmark for evaluating memory and continual learning in LLM systems as alternative to scaling parameters and data.

Ax Jingxin Zhan, Yuze Han, Zhihua Zhang 5/4/2026

Last-Iterate Analyses of FTRL with the 1/2-Tsallis Entropy in Stochastic Bandits

Last-iterate convergence analysis of FTRL algorithm with Tsallis entropy in stochastic multi-armed bandits.

Ax Arthur Chen, Victor Zhong 5/4/2026

SynQuE: Estimating Synthetic Dataset Quality Without Annotations

SynQuE formalizes synthetic dataset quality estimation using proxy metrics to rank datasets without annotations for data-scarce settings.

Ax Gaoxiang Zhao, Ruinan Qiu, Pengpeng Zhao, Rongjin Wang, Xiaoting Wang, Zhangang Lin, Xiaoqiang Wang 5/4/2026

Uncertainty Modeling for Multi-Objective RTA Interception with Distillation Acceleration

Uncertainty modeling approach for Real-Time Auction interception with distillation acceleration for traffic quality estimation.