Ax Kai Pan 5/12/2026

Agent-First Tool API: A Semantic Interface Paradigm for Enterprise AI Agent Systems

Agent-First Tool API paradigm redesigns APIs for autonomous agents, addressing architectural mismatches with human-oriented CRUD APIs.

Ax Adam Gould, Francesca Toni 5/12/2026

Deep Arguing

Examines lack of explicit reasoning and interpretability in deep learning models and proposals for understanding representations.

Ax Shray Mathur, J. Anibal Boscoboinik, Esther H. R. Tsai, Kevin G. Yager 5/12/2026

LLM Jaggedness Unlocks Scientific Creativity

SciAidanBench measures scientific creativity of LLMs through open-ended questions, examining uneven capability progress.

Ax Philipp Steigerwald, Mara Stieler, Jennifer Burghardt, Eric Rudolph, Jens Albrecht 5/12/2026

LLARS: Enabling Domain Expert & Developer Collaboration for LLM Prompting, Generation and Evaluation

LLARS open-source platform enabling collaboration between domain experts and developers for LLM prompt engineering and evaluation.

Ax Maxime Bouscary, Manxi Wu, Saurabh Amin 5/12/2026

Budget-Efficient Automatic Algorithm Design via Code Graph

Budget-efficient automatic algorithm design using LLMs via code graph representation to reuse algorithmic components.

Ax Phongsakon Mark Konrad, Tim Lukas Adam, Ane Cathrine Holst Merrild, Riccardo Terrenzi, Rebecca De Rosa, Toygar Tanyel, Serkan Ayvaz 5/12/2026

The Open-Box Fallacy: Why AI Deployment Needs a Calibrated Verification Regime

Argues for calibrated verification over mechanistic interpretability for AI deployment authorization in sensitive domains.

Ax Riya Tapwal, Abhishek Kumar, Carsten Maple 5/12/2026

PRISM: Generation-Time Detection and Mitigation of Secret Leakage in Multi-Agent LLM Pipelines

PRISM detection system for preventing secret leakage propagation through shared context in multi-agent LLM pipelines.

Ax Ramesh Arvind Naagarajan, Z\"uhal Wagner, Stefan Streif 5/12/2026

Hierarchical Causal Abduction: A Foundation Framework for Explainable Model Predictive Control

Hierarchical causal framework for explainable model predictive control in safety-critical infrastructure.

Ax Minyu Chen, Song Qin, Ling-I Wu, Jianxin Xue, Guoqiang Li 5/12/2026

Teacher-Aware Evolution of Heuristic Programs from Learned Optimization Policies

Evolutionary framework using learned optimization policies as teachers to generate heuristic programs for combinatorial optimization.

Ax Regina Gugg, Selina Niederl\"ander, Andreas St\"ockl, Martin Flechl 5/12/2026

Navigating the Sea of LLM Evaluation: Investigating Bias in Toxicity Benchmarks

Investigation of bias and robustness issues in LLM toxicity evaluation benchmarks used for deployment certification.

Ax Zhiyuan Fan, Wenwei Jin, Feng Zhang, Bin Li, Yihong Dong, Yao Hu, Jiawei Li 5/12/2026

Evolving-RL: End-to-End Optimization of Experience-Driven Self-Evolving Capability within Agents

Framework for self-evolving agents that distill experience from interactions to adapt to novel tasks at deployment time.

Ax Xinrun Wang, Chang Yang, He Zhao, Zhuoyi Lin, Shuyue Hu 5/12/2026

The Agent Use of Agent Beings: Agent Cybernetics Is the Missing Science of Foundation Agents

Theoretical framework examining agent cybernetics as foundational science for long-horizon LLM agents using tool loops and reflection.

Ax Tim Van hamme, Thomas Vissers, Javier Carnerero-Cano, Mario Fritz, Emil C. Lupu, Lieven Desmet, Dinil Mon Divakaran 5/12/2026

MATRA: Modeling the Attack Surface of Agentic AI Systems -- OpenClaw Case Study

MATRA threat modeling framework for assessing security risks in LLM-based agent systems with tool access.

Ax Lihuan Li, Wilson Wongso, Baiyu Chen, Hao Xue, Ruiyi Yang, Yifan Duan, Xiachong Lin, Yang Song, Flora Salim 5/12/2026

TrajPrism: A Multi-Task Benchmark for Language-Grounded Urban Trajectory Understanding

Multi-task benchmark for aligning language descriptions with urban trajectory data, focusing on mobility modeling and planning.

Ax Yuanyang Li, Xue Yang, Longyue Wang, Weihua Luo, Hongyang Chen 5/12/2026

ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox

ComplexMCP benchmark evaluates LLM agents on interdependent tool use in realistic scenarios with 300+ tests on Model Context Protocol.

Ax Shengxiang Gao, Chao Lei, Jey Han Lau, Jianzhong Qi 5/12/2026

PathISE: Learning Informative Path Supervision for Knowledge Graph Question Answering

Knowledge graph QA method using LLMs with retrieval-augmented generation, proposing informative path supervision for training.

Ax Wenbo Zhang, Lijinghua Zhang, Liner Xiang, Hengrui Cai 5/12/2026

Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for LLM-as-a-Judge

Compares reasoning vs non-reasoning LLM judges showing structured verification benefits vary by task, proposes cost-efficient routing for LLM-as-a-Judge applications.

Ax Jinhang Xu, Qiyuan Zhu, Yujun Wu, Zirui Wang, Dongxu Zhang, Jianxin Tang, Marcia Tian, Yiling Duan, Siyuan Li, Jingxuan Wei, Sirui Han, Yike Guo, Odin Zhang, Conghui He, Cheng Tan 5/12/2026

NanoResearch: Co-Evolving Skills, Memory, and Policy for Personalized Research Automation

NanoResearch co-evolves skills, memory, and policy for personalized multi-agent research automation, adapting outputs to individual researcher resource constraints and preferences.

Ax Jihoo Jung, Chaeyoung Jung, Ji-Hoon Kim, Joon Son Chung 5/12/2026

Probing Cross-modal Information Hubs in Audio-Visual LLMs

Investigates internal mechanisms and cross-modal information processing in audio-visual LLMs through interpretability probing techniques. Limited practical application focus.

Ax Moksh Jain, Mehdi Bennani, Johannes Bausch, Yuri Chervonyi, Bogdan Georgiev, Simon Osindero, Nenad Toma\v{s}ev 5/12/2026

MaD Physics: Evaluating information seeking under constraints in physical environments

MaD Physics benchmark evaluates agents for scientific discovery under physical measurement constraints, balancing quality and quantity with cost-benefit trade-offs.

Ax Muhan Gao, Zih-Ching Chen, Kuan-Hao Huang 5/12/2026

The First Drop of Ink: Nonlinear Impact of Misleading Information in Long-Context Reasoning

Studies nonlinear impact of misleading information on LLM long-context performance, quantifying how distractors degrade retrieval-augmented generation and agentic systems.

Ax Pedro Conde, Henrique Branquinho, Valerio Mazzone, Bruno Mendes, Andr\'e Baptista, Nuno Moniz 5/12/2026

From Controlled to the Wild: Evaluation of Pentesting Agents for the Real-World

Evaluates AI pentesting agents on real-world security targets beyond predefined CTF benchmarks, assessing offensive capabilities in unconstrained environments.

Ax Daniel Mitropolsky, Susan S. Hong, Riccardo Neumarker, Emanuele Rimoldi, Tomaso Poggio 5/12/2026

The Generalized Turing Test: A Foundation for Comparing Intelligence

Generalized Turing Test framework for comparing arbitrary agent intelligence through indistinguishability via Turing comparators, dataset and task-agnostic.

Ax Haozhe Zhang, Kaichen Liu, Miaomiao Chen, Lei Li, Shaojie Yang, Cheng Peng, Hanjie Chen 5/12/2026

BenchCAD: A Comprehensive, Industry-Standard Benchmark for Programmatic CAD

BenchCAD comprehensive benchmark for programmatic CAD code generation from visual/textual inputs, requiring understanding of 3D structure, parameters, and engineering operations.

Ax Mingxi Zou, Zhihan Guo, Langzhang Liang, Zhuo Wang, Qifan Wang, Qingsong Wen, Irwin King, Lizhen Qu, Zenglin Xu 5/12/2026

Remember the Decision, Not the Description: A Rate-Distortion Framework for Agent Memory

Rate-distortion framework for agent memory optimization preserving decision-relevant distinctions under limited runtime budgets rather than descriptive relevance criteria.

Ax Simon Yu, Derek Chong, Ananjan Nandi, Dilara Soylu, Jiuding Sun, Christopher D Manning, Weiyan Shi 5/12/2026

Shepherd: A Runtime Substrate Empowering Meta-Agents with a Formalized Execution Trace

Shepherd runtime substrate formalizes meta-agent operations with typed execution traces and Git-like replay, achieving 5x faster process forking than Docker and 95%+ prompt-cache reuse.

Ax Ali Saraeb 5/12/2026

Artificial Intelligence in Number Theory: LLMs for Algorithm Generation and Ensemble Methods for Conjecture Verification

Evaluates open-source LLMs for algorithm generation and ensemble methods for conjecture verification in number theory, demonstrating specialized domain application.

Ax Hao Yu 5/12/2026

A Unified Pair-GRPO Family: From Implicit to Explicit Preference Constraints for Stable and General RL Alignment

Pair-GRPO unified framework addresses instability in LLM RLHF alignment through preference-based RL optimization with improved gradient interpretability and variance reduction.

Ax Mahdi Erfanian, Nelson Daniel Troncoso, Aashna Garg, Amabel Gale, Xiaoyu Liu, Pareesa Ameneh Golnari, Shengyu Fu 5/12/2026

Delulu: A Verified Multi-Lingual Benchmark for Code Hallucination Detection in Fill-in-the-Middle Tasks

Delulu benchmark for detecting code hallucinations in LLM fill-in-the-middle tasks across 7 languages with 1,951 verified samples and 4 hallucination types.

Ax Maribeth Rauh, Dick A. H. Blankvoort, Matias Duran, Caoilfhionn N\'i Dheor\'ain, Harshvardhan J. Pandit, Siddharth D. Jaiswal, Anthony Ventresque, Abeba Birhane 5/12/2026

Playing Games with My Heart: An Evaluation of AI Companion Apps

Research evaluating AI companion chatbots for simulated relationships, analyzing emotional dependence risks and psychological harm. No direct developer or technical focus.

Ax Xinchun Su, Chunxu Luo, Lipeng Ma, Yixuan Li, Weidong Yang 5/12/2026

MedThink: Enhancing Diagnostic Accuracy in Small Models via Teacher-Guided Reasoning Correction

MedThink uses knowledge distillation and teacher-guided reasoning correction to compress LLM diagnostic capabilities into smaller models for resource-constrained clinical settings.

Ax Dario Coscia, Sindy L\"owe, Max Welling 5/12/2026

BaLoRA: Bayesian Low-Rank Adaptation of Large Scale Models

BaLoRA extends LoRA fine-tuning with Bayesian uncertainty quantification for large pre-trained models.

Ax Omar Faruque, Sahara Ali, Xue Zheng, Jianwu Wang 5/12/2026

TTCD:Transformer Integrated Temporal Causal Discovery from Non-Stationary Time Series Data

Transformer-based causal discovery method for non-stationary time series data with contemporary and lagged relationships.

Ax Drew Dillon, Kasyap Varanasi 5/12/2026

Context-Augmented Code Generation: How Product Context Improves AI Coding Agent Decision Compliance by 49%

Research showing product context improves AI coding agent decision compliance by 49%, with benchmark across 8 software engineering tasks.

Ax Amman Yusuf, Zhejun Jiang, Mijung Park 5/12/2026

The Safety-Aware Denoiser for Text Diffusion Models

Safety-Aware Denoiser framework for controlling safety risks in text diffusion models with guidance mechanisms.

Ax Yongzhong Xu 5/12/2026

Feature Repulsion and Spectral Lock-in: An Empirical Study of Two-Layer Network Grokking

Empirical analysis of feature repulsion and spectral properties during two-layer neural network grokking.

Ax Jiaxin Qi, Hang Li, Yan Cui, Yuhua Zheng, Jianqiang Huang 5/12/2026

Towards Universal Gene Regulatory Network Inference: Unlocking Generalizable Regulatory Knowledge in Single-cell Foundation Models

Foundation model approach for gene regulatory network inference from single-cell transcriptomic data.

Ax Rui Zhao, Haofeng Hu, Zhenhai Gao, Jiaqiao Liu, Gao Fei 5/12/2026

VLADriver-RAG: Retrieval-Augmented Vision-Language-Action Models for Autonomous Driving

Retrieval-augmented vision-language-action model for autonomous driving addressing long-tail scenario generalization.

Ax Natalia Frumkin, Bokun Wang, Hung-Yueh Chiang, Chi-Chih Chang, Mohamed S. Abdelfattah, Diana Marculescu 5/12/2026

DARE: Diffusion Language Model Activation Reuse for Efficient Inference

Activation reuse technique exploiting token-wise redundancy in diffusion language models for efficient inference.

Ax Plawan Kumar Rath, Rahul Maliakkal 5/12/2026

Weight Pruning Amplifies Bias: A Multi-Method Study of Compressed LLMs for Edge AI

Empirical study showing weight pruning amplifies bias in compressed LLMs across three models and pruning methods.

Ax Yuan Fang, Yi Xie, Xuming Ran 5/12/2026

HoReN: Normalized Hopfield Retrieval for Large-Scale Sequential Model Editing

Hopfield network-based approach for sequential model editing of LLMs preserving factual knowledge during updates.

Ax Fang Wu, Haokai Zhao, Da Xing, Hanqun Cao, Tinson Xu, Yanchao Li, Xiangru Tang, Zehong Wang, Aaron Tu, Kuan Pang, Hanchen Wang, Hongbin Lin, Zeqi Zhou, Yinxi Li, Peng Xia, Li Erran Li, Molei Tao, Jure Leskovec, Aditya Joshi, Yejin Choi 5/12/2026

NoiseRater: Meta-Learned Noise Valuation for Diffusion Model Training

Meta-learning framework assigning importance scores to noise instances in diffusion model training.

Ax Yuriel Ryan, Hei Man Ip, Adriel Kuek, Paul Pu Liang, Roy Ka-Wei Lee 5/12/2026

Self-Captioning Multimodal Interaction Tuning: Amplifying Exploitable Redundancies for Robust Vision Language Models

Method for improving vision-language model robustness by analyzing multimodal redundancy and synergistic information.

Ax Zi-Yi Jia, Zi-Jian Cheng, Xin-Yue Zhang, Kun-Yang Yu, Zhi Zhou, Yu-Feng Li, Lan-Zhe Guo 5/12/2026

VT-Bench: A Unified Benchmark for Visual-Tabular Multi-Modal Learning

First unified benchmark for visual-tabular multimodal learning across 14 datasets in 9 domains.

Ax Jincheng Xie, Yawen Ling, Qi Xiao, Feiyu Zhang, Zhongyi Huang, Wen Hu, Yu Zheng 5/12/2026

SPECTRE: Hybrid Ordinary-Parallel Speculative Serving for Resource-Efficient LLM Inference

Speculative decoding serving framework for resource-efficient multi-model LLM inference using tail models as drafters.

Ax Divya Gupta 5/12/2026

Privacy-Preserving Federated Learning: Integrating Zero-Knowledge Proofs in Scalable Distributed Architectures

Federated learning framework integrating zero-knowledge proofs for privacy-preserving distributed AI model training.

Ax Haopeng Jin, Hongzhu Yi, Wenlong Zhao, Jinwen Luo, Shani Ye, Zhenyu Guan, Shiquan Dong, Tiankun Yang, Tao Yu 5/12/2026

HY-Himmel Technical Report: Hierarchical Interleaved Multi-stream Motion Encoding for Long Video Understanding

Hierarchical video-language model reducing token growth and improving motion perception for long video understanding.

Ax Rebekah Lane, Logan Cummins, Andy Perkins, George Trawick, Ioana Banicescu, Sudip Mittal 5/12/2026

parHSOM: A novel parallel Hierarchical Self-Organizing Map implementation

Parallel implementation of Hierarchical Self-Organizing Maps for intrusion detection systems in cybersecurity.

Ax Ayoub Agouzoul 5/12/2026

Understanding Asynchronous Inference Methods for Vision-Language-Action Models

Methods for reducing inference latency in Vision-Language-Action models through inpainting, delay simulation, and residual correction techniques.

Ax Lurong Pan 5/12/2026

Communication Dynamics Neural Networks: FFT-Diagonalized Layers for Improved Hessian Conditioning at Reduced Parameter Count

FFT-based neural network layers using circulant spectral machinery for improved Hessian conditioning with fewer parameters.