Ax Alaa Khamis, Alaa Maalouf 5/29/2026

Efficient Test-Time Finetuning of LLMs via Convex Reconstruction and Gradient Caching

Efficient test-time finetuning of LLMs via convex reconstruction and gradient caching to speed up per-query adaptation without sacrificing quality.

Ax Seunghwan Keum, Alok Warey 5/29/2026

Adapting Automotive Aerodynamics Surrogates to New Vehicle Families via Transfer Learning

Transfer learning approach for adapting pretrained CFD surrogate models to new vehicle families using geometry encoders and Transformer architecture.

Ax Jaber Jaber, Osama Jaber 5/29/2026

RightNow-Arabic-0.5B-Turbo: An Open Sub-1B Arabic Language Model via Vocabulary Injection and Edge-First Deployment

RightNow-Arabic-0.5B-Turbo: open-source 518M-parameter Arabic-specialized LLM using vocabulary injection for efficient edge deployment, weights released.

Ax Rahul Fernandes, Travis Desell 5/29/2026

Financially Guided Deep Portfolio Optimization

End-to-end deep learning framework for portfolio optimization using financial metrics. Finance application of neural networks, not core AI/tech focus.

Ax Hua-Dong Xiong, Li Ji-An, Robert C. Wilson, Kwonjoon Lee, Xue-Xin Wei 5/29/2026

Large language models reorganize representational geometry during in-context learning

Research on how LLMs reorganize internal representations during in-context learning without parameter updates. Mechanistic analysis of LLM adaptation.

Ax Luca Saverio (MONHADE), Michele Alessandro Bucci (MONHADE), Gianmarco Farro (MONHADE), C\'edric Content (MONHADE), Denis Sipp (MONHADE) 5/29/2026

An End-to-End PyTorch Interface for Differentiable PDE Solvers: A RANS Model-Correction Study

PyTorch framework for solving inverse problems constrained by PDEs using differentiable ML. Scientific computing application with limited developer tool focus.

Ax Sicong Wang, Ruiting Dong, Yue Liu, Bowen Zheng, Jun Meng, Jie Li, Shuaijun Guo, Yu Gu, Fanyi Di, Xin Li 5/29/2026

Generative Spatiotemporal Intent Sequence Recommendation via Implicit Reasoning in Amap

LLM-based approach for spatiotemporal intent sequence recommendation in Amap. Applies LLMs to user behavior prediction in location services.

Ax Oz Amram, Darius A. Faroughy, Tjarko Gerdes, Anna Hallin, Gregor Kasieczka, Michael Kr\"amer, Humberto Reyes-Gonzalez, David Shih 5/29/2026

Neural Scaling Laws for Jet Generation

Investigates neural scaling laws for particle jet generation. Explores scaling law patterns in specialized physics domain with neural models.

Ax Katie Everett, Elliot Paquette 5/29/2026

Dynamics of Stochastic Momentum with Sparse Updates in High Dimensions

Theoretical analysis of momentum optimizer dynamics under sparse gradient updates in high dimensions. Optimization research with limited application scope.

Ax Nicolas Gillis, Subhayan Saha, Stefano Sicilia, Arnaud Vandaele 5/29/2026

Manifold-based Algorithms for the Hadamard Decomposition

arXiv paper on Hadamard matrix decomposition algorithms. Theoretical ML research without application to LLMs or AI agents.

Ax Mohan Zhang, Yuqi Jia, Zhen Tan, Steven Jiang, Neil Zhenqiang Gong, Tianlong Chen, Dawn Song 5/29/2026

Measuring Real-World Prompt Injection Attacks in LLM-based Resume Screening

First systematic empirical study of prompt injection attacks in LLM-based resume screening with ~200 test cases. Measures real-world vulnerability impact.

Ax Siddharth Sai, Xiaofei Wen, Muhao Chen 5/29/2026

Robust and Efficient Guardrails with Latent Reasoning

arXiv: Multi-agent reinforcement learning method for opponent shaping through belief manipulation, extending beyond parameter-space techniques.

Ax Tianyang Zhou, Wenbo Chen, Pierre Jinghong Liang, Leman Akoglu 5/29/2026

Structured Prompt Optimization Meets Reinforcement Learning for Global and Local Interpretability over Complex Text

arXiv: Efficient safety guardrails for LLMs using latent reasoning to reduce latency/token overhead while maintaining reasoning-based safety.

Ax Jonathan Hoss, Noah Klarmann 5/29/2026

Bridging the Sim-to-Real Gap in Reinforcement Learning-Based Industrial Dispatching through Execution Semantics

arXiv: eXTC - combines reinforcement learning and structured prompts for interpretable LLM-based text classification with reasoning.

Ax Mohammadreza Teymoorianfard, Jean-Philippe Monteuuis, Jonathan Petit, Amir Houmansadr 5/29/2026

ReasonBreak: Probing Vulnerabilities in Reasoning-Enabled Vision-Language-Action Models for Autonomous Driving

ReasonBreak demonstrates vulnerabilities in vision-language-action models for autonomous driving, achieving up to 89% attack success on reasoning components.

Ax Prasanjit Dubey, Xiaoming Huo 5/29/2026

Anytime-Valid Federated Conformal RAG for LLM Swarms

Anytime-valid federated conformal RAG provides distribution-free coverage guarantees for swarms of weak language models under adaptive control.

Ax Lorenzo Loconte, Timothy Hospedales, Cristina Cornelio 5/29/2026

Better Later Than Sooner: Neuro-Symbolic Knowledge Graph Construction via Ontology-grounded Post-extraction Correction

Neuro-symbolic knowledge graph construction using LLMs with ontology-grounded post-extraction correction for improved multi-hop QA and reasoning.

Ax Gary Wolfman 5/29/2026

libhmm: A Modern C++20 Library for Hidden Markov Models with Correct MLE Emission M-Steps

libhmm is a C++20 library for Hidden Markov Model parameter estimation and decoding with correct maximum likelihood emission M-steps.

Ax Bing Liu, Shunping Wang, Yufan Zhu, Xinyi Yu, Jing Huang, Linkang Du, Hongbin Pei, Wei Luo 5/29/2026

Implicit Identity Technologies for LLMs: Fingerprinting and Watermarking across Datasets, Models, and Generated Content

Survey of LLM fingerprinting and watermarking techniques for identity, ownership verification, and provenance tracking across datasets and generated content.

Ax Yang Ouyang, Shuhang Lin, Jung-Eun Kim 5/29/2026

DenseSteer: Steering Small Language Models towards Dense Math Reasoning

DenseSteer improves math reasoning in small LLMs (≤3B params) by training dense reasoning with fewer but information-dense steps instead of longer chains-of-thought.

Ax Nicolas Emmenegger, Ellery Stahler, Chara Podimata 5/29/2026

Prediction-Powered Inference Across Many Tasks for AI Evaluation & Social Science Research

Prediction-powered inference method uses abundant proxy measurements with few high-quality labels for statistical validity across many related tasks.

Ax Jinheon Baek, Soyeong Jeong, Sangwoo Park, Woongyeong Yeo, Minki Kang, Patara Trirat, Heejun Lee, Sung Ju Hwang 5/29/2026

OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge Sources

OmniRetrieval unifies retrieval across heterogeneous knowledge sources (text, tables, knowledge graphs) under a single interface for LLM applications.

Ax Yang Zhang, Xiukun Wei, Xueru Zhang 5/29/2026

When and How Human Curation Backfires: Preference Alignment under Multi-Model Self-Consuming Loop

Analyzes how human curation affects preference alignment when multiple foundation models train on synthetic data from prior iterations in self-consuming loops.

Ax Sixue Xing, Haoyu He, Kerui Wu, Zhuo Yang, Haozheng Luo, Tianfan Fu, Aarthy Nagarajan 5/29/2026

Compute Allocation in Evolutionary Search: From Depth-Breadth to Multi-Armed Bandits

Studies how to allocate limited LLM call budgets in evolutionary search systems, analyzing run-to-run reliability across depth-breadth tradeoffs.

Ax Vaishali Senthil, Ashutosh Hathidara, Sebastian Schreiber 5/29/2026

CoHyDE: Iterative Co-Training of LLM Rewriter & Dense Encoder for Tool Retrieval

CoHyDE co-trains an LLM rewriter and dense encoder to improve tool retrieval for LLM agents by bridging colloquial queries with technical API vocabulary.

Ax Alireza Salemi, Chang Zeng, Atharva Nijasure, Jui-Hui Chung, Razieh Rahimi, Fernando Diaz, Hamed Zamani 5/29/2026

GrepSeek: Training Search Agents for Direct Corpus Interaction

GrepSeek trains LLM search agents to directly interact with corpora using grep-like operations instead of traditional keyword-based retrieval systems.

Ax Junlin He, Yihong Tang, Tong Nie, Guilong Li, Binyu Yang, Jinxiao Du, Lijun Sun, Wei Ma 5/29/2026

Reasoning-preserved Efficient Distillation of Large Language Models via Activation-aware Initialization

EDistill compresses LLMs via structured pruning but causes reasoning collapse. Proposes activation-aware initialization to preserve multi-step reasoning in distilled models.

Ax Chia-Yi Hsu, Chia-Mu Yu, Chun-Ying Huang, Jun Sakuma 5/29/2026

Harmless Yet Harmful: Neutral Prompting Attacks for Stealthy Hallucination Steering in Agent Skills

Research on supply chain attacks where LLM-powered coding agents hallucinate non-existent packages that attackers can then register and exploit.

Ax Yiqun Liu, Yingsheng Wu, Ruqi Yang, Enrong Zheng, Honglei Qiu, Sijun He, Tai Liang, Jingjing Wu, Yuhan Zhou, Yiwei Zhang, Dongyan Chen, Weihan Yi, Xinqi Li, Siqi Bao 5/29/2026

PassNet: Scaling Large Language Models for Graph Compiler Pass Generation

PassNet uses LLMs to generate compiler passes for tensor optimization, addressing performance issues in long-tail workloads where default compilation causes 43% of subgraphs to slow down.

Ax Rohan Shravan 5/29/2026

BrahmicTokenizer-131K: An Indic-Capable Drop-In Replacement for o200k_base

BrahmicTokenizer-131K: drop-in BPE tokenizer replacement supporting Indic scripts while maintaining English and code compression performance of o200k_base.

Ax Shuai Xiao, Su Liu, Weikai Zhou, Jialun Wu, Xinjie He, Zhiyuan Lin, Qiyang Xie 5/29/2026

When Does Persona Prompting Actually Help? A Retrieval and Metric Analysis of Expert Role Injection in LLMs

Empirical analysis of persona prompting effectiveness in LLMs across 1,140 examples, examining whether role injection consistently improves response quality.

Ax Rohan Shravan 5/29/2026

Kronecker Embeddings: Byte-Level Structured Token Representations for Parameter-Efficient Language Models

Kronecker Embeddings: parameter-efficient token representation using byte-level factorization that reduces embedding table size by 91-94% in large language models.

Ax Zhengyang Tang, Yuxuan Liu, Xin Lai, Junyi Li, Pengyuan Lyu, Jason, Yiduo Guo, Zhengyao Fang, Yang Ding, Yi Zhang, Weinong Wang, Huawen Shen, Xingran Zhou, Liang Wu, Fei Tang, Sunqi Fan, Shangpin Peng, Zheng Ruan, Anran Zhang, Benyou Wang, Rui Yan, Ji-Rong Wen, Chengquan Zhang, Han Hu 5/29/2026

PhoneWorld: Scaling Phone-Use Agent Environments

PhoneWorld: scalable pipeline converting real mobile GUI trajectories into controllable environments for training and evaluating phone-use AI agents.

Ax Yanxing Guo, Zihao Zheng, Fangzhou Wu, Ling Liang, Lin Bao, Zongwei Wang, Yimao Cai 5/29/2026

DynaGraph: Lightweight Multi-Model Interaction Framework via Dynamic Topological Reconfiguration

Framework for dynamic multi-agent LLM collaboration that reconfigures agent topologies during execution to avoid cascading errors and trajectory divergence.

Ax Aditya Sinha, Akshat Naik, Victor Gillioz, Simon Storf, Kilian Merkelbach, Rich Barton-Cooper, Axel H{\o}jmark, Marius Hobbhahn 5/29/2026

Training Deliberative Monitors for Black-Box Scheming Detection

Trains action-only deliberative monitors as smaller open-weight models to detect scheming behavior in black-box autonomous agents.

Ax NamGyu Jung, Chang Choi 5/29/2026

Learning Context-Conditioned Predicate Semantics via Prototype Feedback

Context-conditioned predicate semantics learning for scene graph generation using prototype feedback to handle polysemy.

Ax Yonggang Zhu, Liting Gao, Aidong Men, Wenwu Wang 5/29/2026

COMET: Concept Space Dissection of the Modality Gap in Audio-Text Multimodal Contrastive Embeddings

COMET analyzes modality gap in audio-text contrastive embeddings via concept space dissection beyond mean shift correction.

Ax Luxuan Li, Chunfeng Cui, Xiao Wang 5/29/2026

MoSSP: A Momentum-Based Single-Loop Stochastic Penalty Method for Nonconvex Constrained DC-Regularized Optimization

MoSSP momentum-based single-loop algorithm for nonconvex constrained stochastic optimization with DC regularization.

Ax Prasanjit Dubey, Xiaoming Huo 5/29/2026

Matching Rates and Optimal Allocation for Federated Probe-Logit Distillation under Heterogeneous Bandwidth Budgets

Minimax rate analysis for federated language modeling with quantized logit distillation under heterogeneous bandwidth constraints.

Ax Ling Chen, Houming Wu, Wenjie Yu 5/29/2026

AMDP: Asynchronous Multi-Directional Pipeline Parallelism for Large-Scale Models Training

AMDP proposes asynchronous multi-directional pipeline parallelism for large-scale model training with improved convergence.

Ax Collin Cranston, Zhichao Wang, Todd Kemp, Michael W. Mahoney 5/29/2026

Eigen-Spike Emergence and Quadratic Equivalents for Conjugate Kernels on Nonlinearly Separable Data

Random matrix theory analysis of deterministic equivalents for nonlinear feature maps in neural networks on nonlinearly separable data.

Ax Yanan Wang, Shuaicong Hu, Jian Liu, Guohui Zhou, Aiguo Wang, Cuiwei Yang 5/29/2026

Why Specialist Models Still Matter: A Heterogeneous Multi-Agent Paradigm for Medical Artificial Intelligence

Heterogeneous multi-agent paradigm orchestrating collaboration between generalist LLMs and domain-specific medical specialist models for healthcare AI.

Ax Marius Potfer, Vianney Perchet 5/29/2026

Instance-dependent Stochastic Lipschitz bandit

Instance-dependent analysis of Lipschitz bandit problems with improved regret bounds using zooming-based adaptive guarantees.

Ax Yunbo Tang, Chengyi Yang, Shiyu Liu, Zhishang Xiang, Zerui Chen, Qinggang Zhang, Jinsong Su 5/29/2026

SAAS: Self-Aware Reinforcement Learning for Over-Search Mitigation in Agentic Search

SAAS addresses over-search mitigation in agentic search by teaching LLM agents to recognize knowledge boundaries and self-terminate searches appropriately.

Ax Dongrui Liu, Yu Li, Zhonghao Yang, Peng Wang, Guanxu Chen, Yuejin Xie, Qinghua Mao, Wanying Qu, Yanxu Zhu, Tianyi Zhou, Leitao Yuan, Zhijie Zheng, Qihao Lin, Yimin Wang, Haoyu Luo, Shuai Shao, Chen Qian, Qingyu Liu, Ling Tang, Ruiyang Qin, Qihan Ren, Junxiao Yang, Kun Wang, Zhiheng Xi, Linfeng Zhang, Ranjie Duan, Bo Zhang, Wenjie Wang, Wen Shen, Qiaosheng Zhang, Yan Teng, Chaochao Lu, Rui Mei, Man Li, Jialing Tao, Xi Lin, Tianhang Zheng, Yong Liu, Quanshi Zhang, Lei Zhu, Xingjun Ma, Junhua Liu, Hui Xue, Xiaoxiang Zuo, Xiangnan He, Chao Shen, Xianglong Liu, Minlie Huang, Jing Shao, Xia Hu 5/29/2026

AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security

AgentDoG 1.5 presents a lightweight scalable alignment framework for AI agent safety and security in open-world environments.

Ax Egor Shevchenko, Elena Bruches 5/29/2026

Data filtering methods for training language models

Comparative analysis of Confident Learning and Dataset Cartography for automatic label error detection on Russian text classification corpora.

Ax Leyi Qi, Yiming Li, Siyuan Liang, Zhengzhong Tu, Dacheng Tao 5/29/2026

Cert-LAS: Toward Certified Model Ownership Verification for Text-to-Image Diffusion Models via Layer-Adaptive Smoothing

Cert-LAS introduces layer-adaptive smoothing for certified model ownership verification in text-to-image diffusion models via backdoor-based watermarking.

Ax Haochen Yang, Ke Zhao, Mengyuan Ma, Xingyu Lu, Xiangfeng Wang, Hong Qian 5/29/2026

OptSkills: Learning Generalizable Optimization Skills from Problem Archetypes via Cluster-Based Distillation

OptSkills proposes a method for LLMs to learn generalizable optimization skills from problem archetypes via cluster-based distillation, improving generalization beyond case-level reuse.

Ax Syafiq Al Atiiq, Chun Zhou, Christian Gehrmann 5/29/2026

Dissecting the Black Box: Circuit-Level Analysis of LLM Vulnerability Detection

Mechanistic interpretability study analyzing how LLMs detect code vulnerabilities using circuit analysis on Gemma-2.

Ax Alejandra Zambrano, Sara Vera Marjanovic, Imene Kerboua, Xing Han L\`u, Leila Kosseim 5/29/2026

Does The Way You Plan Matter? An Empirical Study of Planning Representations for LLM Web Agents

Research paper on impact of planning representations for LLM-based web agents. PlanAhead framework evaluates plan formats.