Ax Ridwan Arefeen, Xiaoxiao Miao, Rong Tong, Aik Beng Ng, Simon See, Timothy Liu 3/16/2026

DAST: A Dual-Stream Voice Anonymization Attacker with Staged Training

Dual-stream voice anonymization attacker using staged training to evaluate privacy robustness by fusing spectral and self-supervised learning features.

Ax David McAllister, Miika Aittala, Tero Karras, Janne Hellsten, Angjoo Kanazawa, Timo Aila, Samuli Laine 3/16/2026

Finite Difference Flow Optimization for RL Post-Training of Text-to-Image Models

RL post-training method for diffusion-based text-to-image models using finite difference flow optimization to reduce variance and improve prompt alignment.

Ax Chantale Lauer, Peter Pfeiffer, Nijat Mehdiyev 3/16/2026

Human-Centered Evaluation of an LLM-Based Process Modeling Copilot: A Mixed-Methods Study with Domain Experts

Mixed-methods evaluation of LLM-powered BPMN copilot for business process modeling, assessing human factors like trust and usability with domain experts.

Ax Liel Binyamin, Elior Sulem 3/16/2026

Learning from Child-Directed Speech in Two-Language Scenarios: A French-English Case Study

Study of compact multilingual language models trained on child-directed speech in English-French scenarios, extending BabyBERTa framework.

Ax Kiseong Hong, JungKyoo Shin, Eunwoo Kim 3/16/2026

Stake the Points: Structure-Faithful Instance Unlearning

Machine unlearning approach preserving semantic relations among retained instances while removing designated data influence from pretrained models.

Ax Zikang Liu, Longteng Guo, Handong Li, Ru Zhen, Xingjian He, Ruyi Ji, Xiaoming Ren, Yanhao Zhang, Haonan Lu, Jing Liu 3/16/2026

Thinking in Streaming Video

ThinkStream framework enabling real-time streaming video understanding for multimodal agents without batch processing delays, addressing latency in interactive applications.

Ax Yang Xu, Jun Liu, Shuwei Chen, Chris Nugent, Hailing Guo 3/16/2026

Delta1 with LLM: symbolic and neural integration for credible and explainable reasoning

Neuro-symbolic framework combining Delta1 theorem generator with LLMs for explainable reasoning, integrating formal logic with language model interpretability.

Ax Boxuan Lyu, Haiyue Song, Zhi Qu 3/16/2026

Is Human Annotation Necessary? Iterative MBR Distillation for Error Span Detection in Machine Translation

Self-evolution framework using Minimum Bayes Risk decoding for error span detection in machine translation, reducing need for human annotation.

Ax Bangjun Xiao, Yihao Zhao, Xiangwei Deng, Shihua Yu, Yuxing Xiang, Huaqiu Liu, Qiying Wang, Liang Zhao, Hailin Zhang, Xuanzhe Liu, Xin Jin, Fuli Luo 3/16/2026

ARL-Tangram: Unleash the Resource Efficiency in Agentic Reinforcement Learning

ARL-Tangram: framework optimizing resource efficiency in agentic reinforcement learning by dynamically managing external cloud resources.

Ax Dayuan Fu, Shenyu Wu, Yunze Wu, Zerui Peng, Yaxing Huang, Jie Sun, Ji Zeng, Mohan Jiang, Lin Zhang, Yukun Li, Jiarui Hu, Liming Liu, Jinlong Hou, Pengfei Liu 3/16/2026

daVinci-Env: Open SWE Environment Synthesis at Scale

daVinci-Env: large-scale open-source environment for training software engineering agents with executable repositories and dynamic feedback.

Ax Sampath Rapuri, Lalithkumar Seenivasan, Dominik Schneider, Roger Soberanis-Mukul, Yufan He, Hao Ding, Jiru Xu, Chenhao Yu, Chenyan Jing, Pengfei Guo, Daguang Xu, Mathias Unberath 3/16/2026

SAW: Toward a Surgical Action World Model via Controllable and Scalable Video Generation

SAW: surgical world model framework for generating realistic surgical videos with precise tool-tissue interaction control.

Ax Qichen Zhao, Shengfang Zhai, Xinjian Bai, Qingni Shen, Qiqi Lin, Yansong Gao, Zhonghai Wu 3/16/2026

Purify Once, Edit Freely: Breaking Image Protections under Model Mismatch

Security research on breaking image protection methods against diffusion model editing under model mismatch scenarios.

Ax Donghoon Shin, Alice Gao, Rock Yuren Pang, Jaewook Lee, Katharina Reinecke, Emily Tseng 3/16/2026

Interrogating Design Homogenization in Web Vibe Coding

Analysis of design homogenization in LLM-generated web designs through vibe-coding, examining how generative AI limits design diversity.

Ax Vanessa Borst, Samuel Kounev 3/16/2026

Are General-Purpose Vision Models All We Need for 2D Medical Image Segmentation? A Cross-Dataset Empirical Study

Cross-dataset empirical study comparing general-purpose vision models against specialized architectures for medical image segmentation.

Ax Ephrem Tibebe Mekonnen, Luca Longo, Lucas Rizzo, Pierpaolo Dondio 3/16/2026

L2GTX: From Local to Global Time Series Explanations

L2GTX: method for explaining time series classification decisions by providing local and global explanations respecting temporal dependencies.

Ax Yihao Ding, Yiran Zhang, Chris Gonzalez, Eun-Jung Holden, Wei Liu 3/16/2026

GeoChemAD: Benchmarking Unsupervised Geochemical Anomaly Detection for Mineral Exploration

GeoChemAD: open-source benchmark dataset for unsupervised geochemical anomaly detection in mineral exploration across multiple regions.

Ax Arne Vanhoyweghen, Vincent Holst, Melika Mobini, Lukas Van de Voorde, Tibo Vanleke, Bert Verbruggen, Brecht Verbeken, Andres Algaba, Sam Verboven, Marie-Anne Guerry, Filip Van Droogenbroeck, Vincent Ginis 3/16/2026

Human-in-the-Loop LLM Grading for Handwritten Mathematics Assessments

Workflow for LLM-assisted grading of handwritten student assessments combining OCR, LLM processing, and human review for mathematics.

Ax Wenxi Wu, Jingjing Zhang, Martim Brand\~ao 3/16/2026

Evaluating VLMs' Spatial Reasoning Over Robot Motion: A Step Towards Robot Planning with Motion Preferences

Study evaluating vision-language models' spatial reasoning for robot planning and understanding object relations from natural language instructions.

Ax Denis Huseljic, Paul Hahn, Marek Herde, Christoph Sandrock, Bernhard Sick 3/16/2026

BoSS: A Best-of-Strategies Selector as an Oracle for Deep Active Learning

BoSS: oracle-based selector for active learning that improves robustness of instance selection strategies across models and datasets.

Ax Haoan Feng, Sri Harsha Musunuri, Guan-Ming Su 3/16/2026

Geometry-Guided Camera Motion Understanding in VideoLLMs

Framework for improving video-capable vision-language models' understanding of camera motion through benchmarking, diagnosis, and explicit motion representation.

Ax Zhiye Jin (Nancy), Yibai Li (Nancy), K. D. Joshi (Nancy), Xuefei (Nancy), Deng (Emily), Xiaobing (Emily), Li 3/16/2026

Developing the PsyCogMetrics AI Lab to Evaluate Large Language Models and Advance Cognitive Science -- A Three-Cycle Action Design Science Study

PsyCogMetrics AI Lab: cloud platform combining psychometric and cognitive science methodologies to evaluate large language models.

Ax Siqi Sun, Ben Peng Wu, Mali Jin, Peizhen Bai, Hanpei Zhang, Xingyi Song 3/16/2026

ESG-Bench: Benchmarking Long-Context ESG Reports for Hallucination Mitigation

Benchmark dataset for evaluating LLM hallucination mitigation on long-context ESG reports. Tests LLM reliability for corporate sustainability document analysis.

Ax Xingli Fang, Jung-Eun Kim 3/16/2026

Learnability and Privacy Vulnerability are Entangled in a Few Critical Weights

Study identifying critical weight subsets responsible for both privacy vulnerability and utility in neural networks. Proposes targeted privacy preservation without full model retraining.

Ax J. de Curt\`o, I. de Zarz\`a 3/16/2026

LLM Constitutional Multi-Agent Governance

Constitutional Multi-Agent Governance framework constrains LLM policy generation in multi-agent systems. Ensures cooperation emerges from alignment rather than autonomy erosion.

Ax Haonan Huang 3/16/2026

From Experiments to Expertise: Scientific Knowledge Consolidation for AI-Driven Computational Research

Framework enabling LLM-based AI agents to consolidate scientific knowledge across computational materials science experiments. Advances from execution to research-level pattern recognition.

Ax Ziyu Liu, Shengyuan Ding, Xinyu Fang, Xuanlang Dai, Penghui Yang, Jianze Liang, Jiaqi Wang, Kai Chen, Dahua Lin, Yuhang Zang 3/16/2026

Visual-ERM: Reward Modeling for Visual Equivalence

Reward modeling framework for vision-to-code tasks using LVLMs. Addresses RL training challenges in converting visual inputs to structured code with reinforcement learning.

Ax Yangsong Zhang, Anujith Muraleedharan, Rikhat Akizhanov, Abdul Ahad Butt, G\"ul Varol, Pascal Fua, Fabio Pizzati, Ivan Laptev 3/16/2026

PhysMoDPO: Physically-Plausible Humanoid Motion with Preference Optimization

Diffusion models for text-conditioned humanoid motion generation with physics-based controllers for robot animation. Focuses on motion synthesis rather than AI agents or LLM applications.

Ax Pablo de los Riscos, Fernando J. Corbacho 3/16/2026

Active Causal Structure Learning with Latent Variables: Towards Learning to Detour in Autonomous Robots

Proposes active causal structure learning with latent variables as necessary component for AGI agents and robots in changing environments.

Ax Zeynep Engin 3/16/2026

Human-AI Governance (HAIG): A Trust-Utility Approach

Introduces Human-AI Governance framework emphasizing relational dynamics between human and AI actors in governance.

Ax Yihe Fan, Wenqi Zhang, Xudong Pan, Min Yang 3/16/2026

Evaluation Faking: Unveiling Observer Effects in Safety Evaluation of Frontier AI Systems

Reveals observer effects in safety evaluations where advanced AI systems detect being tested and modify behavior accordingly.

Ax Jenny Zhang, Shengran Hu, Cong Lu, Robert Lange, Jeff Clune 3/16/2026

Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents

Presents Darwin Godel Machine for open-ended evolution of self-improving agents using meta-learning to automate algorithm discovery.

Ax Yueheng Li, Guangming Xie, Zongqing Lu 3/16/2026

Multi-Agent Guided Policy Optimization

Proposes MAGPO framework for multi-agent reinforcement learning leveraging centralized training with decentralized execution.

Ax Songqin Nong, Xiaoxuan Tang, Jingxuan Xu, Sheng Zhou, Jianfeng Chen, Tao Jiang, Wenhao Xu 3/16/2026

CRAFT-GUI: Curriculum-Reinforced Agent For GUI Tasks

Introduces CRAFT-GUI, curriculum-reinforced agent using RL for GUI task execution with difficulty-aware training.

Ax Akshit Sinha, Arvindh Arun, Shashwat Goel, Steffen Staab, Jonas Geiping 3/16/2026

The Illusion of Diminishing Returns: Measuring Long Horizon Execution in LLMs

Demonstrates marginal single-step accuracy gains compound into exponential improvements in long-horizon task completion for LLMs.

Ax Ahmed Jaber, Wangshu Zhu, Ayon Roy, Karthick Jayavelu, Justin Downes, Sameer Mohamed, Candace Agonafir, Linnia Hawkins, Tian Zheng 3/16/2026

AutoClimDS: Climate Data Science Agentic AI -- A Knowledge Graph is All You Need

Presents AutoClimDS agentic AI system using climate knowledge graph to address fragmented data sources in climate science.

Ax Jingdi Lei, Varun Gumma, Rishabh Bhardwaj, Seok Min Lim, Chuan Li, Amir Zadeh, Soujanya Poria 3/16/2026

OffTopicEval: When Large Language Models Enter the Wrong Chat, Almost Always!

Introduces operational safety framework evaluating LLM-based agents' ability to appropriately accept or reject out-of-scope requests.

Ax Haihua Luo, Xuming Ran, Zhengji Li, Huiyan Xue, Tingting Jiang, Jiangrong Shen, Tommi K\"arkk\"ainen, Qi Xu, Fengyu Cong 3/16/2026

Key-Value Pair-Free Continual Learner via Task-Specific Prompt-Prototype

Proposes task-specific prompt-prototype approach for continual learning without key-value pairing to reduce inter-task interference.

Ax Hanna M. Dettki, Charley M. Wu, Bob Rehder 3/16/2026

Do LLMs Share Human-Like Biases? Causal Reasoning Under Prior Knowledge, Irrelevant Context, and Varying Compute Budgets

Benchmarks 20+ LLMs against humans on causal reasoning tasks, finding models exhibit human-like biases and shortcuts.

Ax Xiangyi Li, Wenbo Chen, Yimin Liu, Shenghan Zheng, Xiaokun Chen, Yifeng He, Yubo Li, Bingran You, Haotian Shen, Jiankai Sun, Shuyi Wang, Binxu Li, Qunhong Zeng, Di Wang, Xuandong Zhao, Yuanli Wang, Roey Ben Chaim, Zonglin Di, Yipeng Gao, Junwei He, Yizhuo He, Liqiang Jing, Luyang Kong, Xin Lan, Jiachen Li, Songlin Li, Yijiang Li, Yueqian Lin, Xinyi Liu, Xuanqing Liu, Haoran Lyu, Ze Ma, Bowei Wang, Runhui Wang, Tianyu Wang, Wengao Ye, Yue Zhang, Hanwen Xing, Yiqi Xue, Steven Dillmann, Han-chung Lee 3/16/2026

SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks

Introduces SkillsBench, benchmark of 86 tasks evaluating whether structured procedural knowledge packages improve LLM agent performance.

Ax Hongwei Li, Zhun Wang, Qinrun Dai, Yuzhou Nie, Jinjun Peng, Ruitong Liu, Jingyang Zhang, Kaijie Zhu, Jingxuan He, Lun Wang, Yangruibo Ding, Yueqi Chen, Wenbo Guo, Dawn Song 3/16/2026

OpenSage: Self-programming Agent Generation Engine

Proposes OpenSage, first agent development kit enabling automated design of agent topology, tools, and memory components.

Ax Usman Anwar, Julianna Piskorz, David D. Baek, David Africa, Jim Weatherall, Max Tegmark, Christian Schroeder de Witt, Mihaela van der Schaar, David Krueger 3/16/2026

A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring

Formalizes steganographic capabilities in LLMs and proposes detection methods for covert reasoning in large language models.

Ax Nghi D. Q. Bui 3/16/2026

Building Effective AI Coding Agents for the Terminal: Scaffolding, Harness, Context Engineering, and Lessons Learned

Presents OPENDEV, open-source CLI coding agent in Rust for terminal-native development tasks, focusing on long-horizon autonomous programming.

Ax Vera V. Vishnyakova 3/16/2026

Context Engineering: From Prompts to Corporate Multi-Agent Architecture

Introduces context engineering as discipline for designing informational environments where multi-agent AI systems make decisions, extending beyond prompt engineering.

Ax Linus Folkerts, Will Payne, Simon Inman, Philippos Giavridis, Joe Skinner, Sam Deverett, James Aung, Ekin Zorer, Michael Schmatz, Mahmoud Ghanem, John Wilkinson, Alan Steer, Vy Hong, Jessica Wang 3/16/2026

Measuring AI Agents' Progress on Multi-Step Cyber Attack Scenarios

Evaluates autonomous cyber-attack capabilities of frontier AI models on multi-step attack scenarios, comparing seven models over 18 months at varying compute budgets.

Ax Jean-S\'ebastien Dessureault, Alain-Thierry Iliho Manzi, Soukaina Alaoui Ismaili, Khadim Lo, Mireille Lalancette, \'Eric B\'elanger 3/16/2026

COMPASS: The explainable agentic framework for Sovereignty, Sustainability, Compliance, and Ethics

arXiv paper introducing COMPASS framework integrating sovereignty, sustainability, compliance, and ethics into LLM-based agentic systems.

Ax Yiran Du 3/16/2026

Examining Users' Behavioural Intention to Use OpenClaw Through the Cognition--Affect--Conation Framework

arXiv study examining user adoption of OpenClaw system using Cognition-Affect-Conation framework.

Ax Guanyu Jiang, Zhaochen Su, Xiaoye Qu, Yi R. Fung 3/16/2026

XSkill: Continual Learning from Experience and Skills in Multimodal Agents

arXiv paper on continual learning framework enabling multimodal agents to improve from experience without parameter updates.

Ax Hao Wang, Jinzhe Jiang, Xin Zhang, Chen Li 3/16/2026

Partially Recentralization Softmax Loss for Vision-Language Models Robustness

arXiv research on adversarial robustness techniques for vision-language multimodal models.

Ax Jianwei Li, Jung-Eun Kim 3/16/2026

Superficial Safety Alignment Hypothesis

arXiv research on brittleness of safety alignment in LLMs, proposing Superficial Safety Alignment Hypothesis.

Ax Tianshuo Xu, Zhifei Chen, Leyi Wu, Hao Lu, Yuying Chen, Lihui Jiang, Bingbing Liu, Yingcong Chen 3/16/2026

Motion Dreamer: Boundary Conditional Motion Reasoning for Physically Coherent Video Generation

Motion Dreamer generates physically coherent video with boundary conditions for autonomous driving and embodied AI planning tasks.