Ax Afonso Simpl\'icio, Gon\c{c}alo Vinagre, Miguel Moura Ramos, Diogo Tavares, Rafael Ferreira, Giuseppe Attanasio, Duarte M. Alves, In\^es Calvo, In\^es Vieira, Rui Guerra, James Furtado, Beatriz Canaverde, Iago Paulo, Vasco Ramos, Diogo Gl\'oria-Silva, Miguel Faria, Marcos Treviso, Daniel Gomes, Pedro Gomes, David Semedo, Andr\'e Martins, Jo\~ao Magalh\~aes 3/30/2026

AMALIA Technical Report: A Fully Open Source Large Language Model for European Portuguese

AMALIA: fully open-source LLM trained on high-quality European Portuguese data with native evaluation benchmark.

Ax Guangzhao Yang, Yu Pan, Shi Qiu, Ningjie Bai 3/30/2026

JAL-Turn: Joint Acoustic-Linguistic Modeling for Real-Time and Robust Turn-Taking Detection in Full-Duplex Spoken Dialogue Systems

JAL-Turn: joint acoustic-linguistic model for turn-taking detection in real-time voice AI agent systems.

Ax In\^es Vieira, In\^es Calvo, Iago Paulo, James Furtado, Rafael Ferreira, Diogo Tavares, Diogo Gl\'oria-Silva, David Semedo, Jo\~ao Magalh\~aes 3/30/2026

ALBA: A European Portuguese Benchmark for Evaluating Language and Linguistic Dimensions in Generative LLMs

ALBA benchmark for evaluating European Portuguese language understanding in generative LLMs, addressing underrepresented language evaluation.

Ax James A. Michaelov, Catherine Arnett, Tyler A. Chang, Pamela D. Rivi\`ere, Samuel M. Taylor, Cameron R. Jones, Sean Trott, Roger P. Levy, Benjamin K. Bergen, Micah Altman 3/30/2026

How Open Must Language Models be to Enable Reliable Scientific Inference?

Analysis of how open vs closed language models impact scientific inference reliability and reproducibility.

Ax Moritz Nottebaum, Matteo Dunnhofer, Christian Micheloni 3/30/2026

Beyond MACs: Hardware Efficient Architecture Design for Vision Backbones

Analysis of efficiency metrics for vision backbone networks, demonstrating MACs limitations on edge devices versus actual execution time.

Ax Juan Gabriel Kostelec, Xiang Wang, Axel Laborieux, Christos Sourmpis, Qinghai Guo 3/30/2026

When Perplexity Lies: Generation-Focused Distillation of Hybrid Sequence Models

Study on knowledge distillation of Transformers into efficient hybrid models with focus on generation quality beyond perplexity metrics.

Ax Yoseph Berhanu Alebachew, Hunter Leary, Swanand Vaishampayan, Chris Brown 3/30/2026

Beyond Code Snippets: Benchmarking LLMs on Repository-Level Question Answering

StackRepoQA: first repository-level benchmark evaluating LLMs on multi-file code question answering and program comprehension.

Ax Eziyo Ehsani, Luca Giamattei, Ivano Malavolta, Roberto Pietrantuono 3/30/2026

Sustainability Is Not Linear: Quantifying Performance, Energy, and Privacy Trade-offs in On-Device Intelligence

Framework profiling energy, latency, and quality trade-offs of running LLMs on edge devices with memory and thermal constraints.

Ax Shihua Zhang, Qiuhong Shen, Shizun Wang, Tianbo Pan, Xinchao Wang 3/30/2026

Make Geometry Matter for Spatial Reasoning

Study on improving vision-language models' spatial reasoning by injecting geometry tokens from 3D foundation models.

Ax Zehai He, Wenyi Hong, Zhen Yang, Ziyang Pan, Mingdao Liu, Xiaotao Gu, Jie Tang 3/30/2026

Vision2Web: A Hierarchical Benchmark for Visual Website Development with Agent Verification

Vision2Web: hierarchical benchmark for evaluating AI agents on website development tasks from UI-to-code generation through full-stack development.

Ax Shaoxuan Li, Zhixuan Zhao, Hanze Deng, Zirun Ma, Shulin Tian, Zuyan Liu, Yushi Hu, Haoning Wu, Yuhao Dong, Benlin Liu, Ziwei Liu, Ranjay Krishna 3/30/2026

PerceptionComp: A Video Benchmark for Complex Perception-Centric Reasoning

PerceptionComp: benchmark for complex long-horizon perception-centric video reasoning requiring compositional temporal evidence and multi-subtask integration.

Ax Xinqi (Lucas), Liu, Ruoxi Hu, Alejandro Ojeda Olarte, Zhuoran Chen, Kenny Ma, Charles Cheng Ji, Lerrel Pinto, Raunaq Bhirangi, Irmak Guzey 3/30/2026

Ruka-v2: Tendon Driven Open-Source Dexterous Hand with Wrist and Abduction for Robot Learning

Ruka-v2: open-source tendon-driven dexterous humanoid hand with 11 DOF, data-driven finger control, buildable under $1,300.

Ax Stephen Wissow, Masataro Asai 3/30/2026

Scale-Adaptive Balancing of Exploration and Exploitation in Classical Planning

Scale-adaptive exploration-exploitation balancing in classical planning using Multi-Armed Bandit theory for improved MCTS-based planners.

Ax Masataro Asai, Stephen Wissow 3/30/2026

Extreme Value Monte Carlo Tree Search for Classical Planning

Extreme value Monte Carlo Tree Search improves unbounded cost-to-go estimation in classical planning, outperforming UCB1-based approaches.

Ax Zilong Wang, Nan Chen, Luna K. Qiu, Ling Yue, Geli Guo, Yang Ou, Shiqi Jiang, Yuqing Yang, Lili Qiu 3/30/2026

ReMe: Scaffolding Personalized Cognitive Training via Controllable LLM-Mediated Conversations

ReMe: LLM-mediated conversational framework for scalable personalized cognitive training with controllable task structure and natural interaction.

Ax Priya T. V., Shrisha Rao 3/30/2026

Deontic Temporal Logic for Formal Verification of AI Ethics

Deontic temporal logic formalization for specifying and verifying ethical behavior in AI systems through formal methods.

Ax Haoyu Wang, Christopher M. Poskitt, Jiali Wei, Jun Sun 3/30/2026

ProbGuard: Probabilistic Runtime Monitoring for LLM Agent Safety

ProbGuard: probabilistic runtime monitoring framework for LLM agent safety enabling proactive risk detection during execution in robotics/automation.

Ax Sijia Liu, Niklas Muennighoff, Kawin Ethayarajh 3/30/2026

Humanline: Online Alignment as Perceptual Loss

Online alignment (GRPO) outperforms offline alignment (DPO) through better approximation of human-perceived model output distribution based on prospect theory.

Ax Yunlong Deng, Boyang Sun, Yan Li, Lingjing Kong, Zeyu Tang, Kun Zhang, Guangyi Chen 3/30/2026

Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens

Causal perspective on reasoning tasks showing selection and reflection mechanisms improve LLM reasoning reliability through self-refinement approaches.

Ax Zhengru Fang, Yu Guo, Yuang Zhang, Haonan An, Wenbo Ding, Yuguang Fang 3/30/2026

Shared Spatial Memory Through Predictive Coding

Multi-agent predictive coding framework for constructing shared spatial memory through mutual uncertainty minimization with information bottleneck objective.

Ax Souradip Poddar, Chia-Tung Ho, Ziming Wei, Weidong Cao, Haoxing Ren, David Z. Pan 3/30/2026

HeaRT: A Hierarchical Circuit Reasoning Tree-Based Agentic Framework for AMS Design Optimization

HeaRT: hierarchical circuit reasoning agentic framework for analog/mixed-signal design optimization with adaptive mechanisms and improved transferability.

Ax Jua Han, Jaeyoon Seo, Jungbin Min, Sieun Choi, Huichan Seo, Jihie Kim, Jean Oh 3/30/2026

Before We Trust Them: Decision-Making Failures in Navigation of Foundation Models

Analysis of decision-making failures in foundation models for navigation across reasoning tasks with complete/incomplete spatial and safety-relevant information.

Ax Yupeng Huo, Yaxi Lu, Zhong Zhang, Haotian Chen, Yankai Lin 3/30/2026

AtomMem : Learnable Dynamic Agentic Memory with Atomic Memory Operation

AtomMem: learnable dynamic memory framework for agents using atomic operations, replacing static hand-crafted workflows for long-horizon problem solving.

Ax Yibo Yang, Fei Lei, Yixuan Sun, Yantao Zeng, Chengguang Lv, Jiancao Hong, Jiaojiao Tian, Tianyu Qiu, Xin Wang, Yanbing Chen, Yanjie Li, Zheng Pan, Xiaochen Zhou, Guanzhou Chen, Haoran Lv, Yuning Xu, Yue Ou, Haodong Liu, Shiqi He, Anya Jia, Yulei Xin, Huan Wu, Liang Liu, Jiaye Ge, Jianxin Dong, Dahua Lin, Wenxiu Sun 3/30/2026

AIDABench: AI Data Analytics Benchmark

AIDABench: comprehensive benchmark for evaluating AI-driven document understanding and processing tools in end-to-end real-world scenarios.

Ax Zhiyu Ni, Zheng Liang, Liangcheng Song, Chenrui Cao, Xian Zhang, Alberto Sangiovanni-Vincentelli, Pierluigi Nuzzo 3/30/2026

Draft-and-Prune: Improving the Reliability of Auto-formalization for Logical Reasoning

Draft-and-Prune method improves auto-formalization reliability by reducing semantic failures when translating natural language to solver-executable logical programs.

Ax Julia Jose, Meghna Manoj Nair, Rachel Greenstadt 3/30/2026

Large-Scale Analysis of Persuasive Content on Moltbook

NLP study detecting political propaganda on Moltbook platform using LLM-based classifiers on 673k posts; analyzes prevalence and concentration patterns.

Ax Steven Johnson 3/30/2026

Governance-Aware Vector Subscriptions for Multi-Agent Knowledge Ecosystems

Governance-aware vector subscriptions for multi-agent systems enabling secure real-time knowledge monitoring with data handling policy enforcement.

Ax Yenchia Feng, Chirag Sharma, Karime Maamari 3/30/2026

Environment Maps: Structured Environmental Representations for Long-Horizon Agents

Environment Maps: persistent agent-agnostic representation to reduce cascading errors in long-horizon LLM-based automation tasks across dynamic interfaces.

Ax Umair Siddique 3/30/2026

The Competence Shadow: Theory and Bounds of AI Assistance in Safety Engineering

Formal framework analyzing whether AI assistants improve or introduce blind spots in safety engineering workflows for physical AI systems.

Ax Zachary Bastiani, Robert M. Kirby, Jacob Hochhalter, Shandian Zhe 3/30/2026

Complexity-Aware Deep Symbolic Regression with Robust Risk-Seeking Policy Gradients

Research paper proposing deep symbolic regression approach using policy gradients for discovering interpretable mathematical expressions from data.

Ax G Abarajithan, Zhenghua Ma, Ravidu Munasinghe, Francesco Restuccia, Ryan Kastner 3/30/2026

CGRA4ML: A Hardware/Software Framework to Implement Neural Networks for Scientific Edge Computing

Open-source CGRA4ML framework for implementing neural networks on configurable hardware for near-sensor scientific edge computing.

Ax Dianwei Chen, Zifan Zhang, Lei Cheng, Yuchen Liu, Xianfeng Terry Yang 3/30/2026

INSIGHT: Enhancing Autonomous Driving Safety through Vision-Language Models on Context-Aware Hazard Detection and Edge Case Evaluation

Research paper proposing INSIGHT framework using vision-language models for hazard detection and edge case evaluation in autonomous driving.

Ax Haodi Xu, Joshua Fan, Feng Tao, Lifen Jiang, Fengqi You, Benjamin Z. Houlton, Ying Sun, Carla P. Gomes, Yiqi Luo 3/30/2026

Biogeochemistry-Informed Neural Network (BINN) for Improving Accuracy of Model Prediction and Scientific Understanding of Soil Organic Carbon

Research paper proposing Biogeochemistry-Informed Neural Networks integrating domain knowledge with ML for soil carbon modeling.

Ax Ghazal Kaviani, Yavuz Yarici, Seulgi Kim, Mohit Prabhushankar, Ghassan AlRegib, Mashhour Solh, Ameya Patil 3/30/2026

Hierarchical and Multimodal Data for Daily Activity Understanding

DARai: multimodal hierarchically-annotated dataset of 200+ hours of daily human activities across 10 environments with 20 sensor types.

Ax Dong Liu, Yanxuan Yu, Jiayi Zhang, Yifan Li, Ben Lengerich, Ying Nian Wu 3/30/2026

FastCache: Fast Caching for Diffusion Transformer Through Learnable Linear Approximation

Research paper proposing FastCache framework for accelerating Diffusion Transformer inference through learnable linear approximation and token caching.

Ax Luca L. Weishaupt, Chengkuan Chen, Drew F. K. Williamson, Richard J. Chen, Guillaume Jaume, Tong Ding, Bowen Chen, Anurag Vaidya, Long Phi Le, Guillaume Jaume, Ming Y. Lu, Faisal Mahmood 3/30/2026

Evidence-based diagnostic reasoning with multi-agent copilot for human pathology

Multi-agent copilot using multimodal LLMs for evidence-based diagnostic reasoning in digital pathology.

Ax Akio Kodaira, Tingbo Hou, Ji Hou, Markos Georgopoulos, Felix Juefei-Xu, Masayoshi Tomizuka, Yue Zhao 3/30/2026

StreamDiT: Real-Time Streaming Text-to-Video Generation

StreamDiT enables real-time streaming text-to-video generation using transformer-based diffusion models.

Ax Ruheng Wang, Hang Zhang, Trieu Nguyen, Shasha Feng, Hao-Wei Pang, Xiang Yu, Li Xiao, Peter Zhiping Zhang 3/30/2026

PepThink-R1: LLM for Interpretable Cyclic Peptide Optimization with CoT SFT and Reinforcement Learning

LLM-based framework using chain-of-thought and reinforcement learning for interpretable cyclic peptide design.

Ax Jiayu Ding, Xinpeng Liu, Zhiyi Pan, Shiqiang Long, Ge Li 3/30/2026

ExtrinSplat: Decoupling Geometry and Semantics for Open-Vocabulary Understanding in 3D Gaussian Splatting

3D Gaussian Splatting method for open-vocabulary scene understanding by decoupling geometry and semantics.

Ax Hongxiang Zhang, Yuan Tian, Tianyi Zhang 3/30/2026

Attention-Aligned Reasoning for Large Language Models

arXiv research on ATAR method improving LLM reasoning by aligning attention with reasoning structure. Prevents critical steps from being buried in extended reasoning chains.

Ax T\'eo Guichoux, Th\'eodor Lemerle, Shivam Mehta, Jonas Beskow, Gustav Eje Henter, Laure Soulier, Catherine Pelachaud, Nicolas Obin 3/30/2026

Gelina: Unified Speech and Gesture Synthesis via Interleaved Token Prediction

arXiv research on unified speech and gesture synthesis via interleaved token prediction. Jointly generates synchronized co-speech gestures and speech from text with discrete autoregressive model.

Ax Ilona van der Linden, Sahana Kumar, Arnav Dixit, Aadi Sudan, Smruthi Danda, David C. Anastasiu, Kai Lukoff 3/30/2026

Generating the Modal Worker: A Cross-Model Audit of Race and Gender in LLM-Generated Personas Across 41 Occupations

arXiv research auditing racial and gender biases in LLM-generated occupational personas across 41 professions. Analyzes 1.5M personas from GPT-4, Gemini, DeepSeek, Mistral against BLS data.

Ax Minsuk Ji, Sanghyeok Lee, Namhyuk Ahn 3/30/2026

Compositional Image Synthesis with Inference-Time Scaling

arXiv research on training-free compositional image synthesis combining object-centric approaches with self-refinement. Uses LLMs to improve layout faithfulness in text-to-image generation.

Ax Shijie Zhou, Viet Dac Lai, Hao Tan, Jihyung Kil, Wanrong Zhu, Changyou Chen, Ruiyi Zhang 3/30/2026

GUI-AIMA: Aligning Intrinsic Multimodal Attention with a Context Anchor for GUI Grounding

arXiv research on GUI grounding for computer-use agents. Proposes intrinsic multimodal attention alignment for mapping natural language to screen regions instead of coordinate generation.

Ax Tiansheng Wen, Yifei Wang, Aosong Feng, Long Ma, Xinyang Liu, Yifan Wang, Lixuan Guo, Bo Chen, Stefanie Jegelka, Chenyu You 3/30/2026

Route Experts by Sequence, not by Token

arXiv research on Sequence-level TopK routing for Mixture-of-Experts LLMs. Minimal modification enabling adaptive expert routing based on token complexity without retraining.

Ax Rongbin Hu, Jeffrey Liu 3/30/2026

Binary Verification for Zero-Shot Vision

arXiv research on training-free binary verification workflow for zero-shot vision using VLMs. Converts open-ended queries to multiple-choice with deterministic resolution.

Ax Hao Li, Qiao Sun 3/30/2026

Any4D: Open-Prompt 4D Generation from Natural Language and Images

arXiv research on 4D generation from natural language and images using embodied world models. Addresses data scarcity and long-horizon video generation challenges.

Ax Zhenchao Tang, Fang Wang, Haohuai He, Jiale Zhou, Tianxu Lv, Jun Zhu, Shouzhi Chen, Minghao Yang, Yu Wang, Jiayang Wu, Yidong Song, Yaokun Li, Jiehui Huang, Dawei Huang, Zhi Song, Jianhua Yao 3/30/2026

Aligning LLMs with Biomedical Knowledge using Balanced Fine-Tuning

arXiv research proposing Balanced Fine-Tuning method for aligning LLMs with biomedical knowledge. Combines SFT and RL using confidence-weighted token optimization for scientific understanding.

Ax Daeun Lee, Subhojyoti Mukherjee, Branislav Kveton, Ryan A. Rossi, Viet Dac Lai, Seunghyun Yoon, Trung Bui, Franck Dernoncourt, Mohit Bansal 3/30/2026

StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos

arXiv research on streaming video understanding with gaze signal interpretation for AR applications. Evaluates multimodal LLMs on temporal reasoning with human attention signals.

Ax Woongyeong Yeo, Kangsan Kim, Jaehong Yoon, Sung Ju Hwang 3/30/2026

WorldMM: Dynamic Multimodal Memory Agent for Long Video Reasoning

arXiv research on multimodal memory architecture for long-form video understanding. Addresses context capacity and visual detail retention in hours-long videos using dynamic memory mechanisms.