Ax Jieming Bian, Lei Wang, Letian Zhang, Jie Xu 3/17/2026

FedTreeLoRA: Reconciling Statistical and Functional Heterogeneity in Federated LoRA Fine-Tuning

FedTreeLoRA addresses statistical and functional heterogeneity in federated LoRA fine-tuning for privacy-preserving LLM adaptation.

Ax Angelika Romanou, Mark Ibrahim, Candace Ross, Chantal Shaib, Kerem Okta, Sam Bell, Elia Ovalle, Jesse Dodge, Antoine Bosselut, Koustuv Sinha, Adina Williams 3/17/2026

Brittlebench: Quantifying LLM robustness via prompt sensitivity

Brittlebench introduces evaluation framework measuring LLM robustness to prompt variations, typos, and real-world input noise.

Ax Valerie B\"urger, Marlie Besouw, Jana Fehr, Riana Minocher, Emma Moorhead, Isabel Velarde, Louis Agha-Mir-Salim, Julia Amann, Alexandra Bannach-Brown, David B. Blumenthal, Kaitlyn Hair, Bert Heinrichs, Moritz Herrmann, Elizabeth Hofvenschi\"old, Sune Holm, Anne A. H. de Hond, Sara Kijewski, Stuart McLennan, Timo Minssen, Marco S. Nobile, Nico Pfeifer, Jessica L. Rohmann, Tony Ross-Hellauer, Marija Slavkovik, Karin Tafur, Eleonora Vigan\`o, Magnus Westerlund, Tracey Weissgerber, Vince I. Madai 3/17/2026

How Meta-research Can Pave the Road Towards Trustworthy AI In Healthcare: Catalogue of Ideas and Roadmap for Future Research

Meta-research examines trustworthy AI in healthcare through interdisciplinary collaboration on transparency, robustness, and evidence standards.

Ax Anirudh Jaidev Mahesh, Ben Griffin, Fuat Alican, Joseph Ternasky, Zakari Salifu, Kelvin Amoaba, Yagiz Ihlamur, Aaron Ontoyin Yin, Aikins Laryea, Afriyie Samuel, Yigit Ihlamur 3/17/2026

From Stochastic Answers to Verifiable Reasoning: Interpretable Decision-Making with LLM-Generated Code

Framework reframes LLMs as code generators for interpretable, verifiable decision-making in high-stakes applications with reproducible reasoning.

Ax Yingsheng Geng, Yuchong Gao, Weihong Wu, Guyue Liu, Jiang Liu 3/17/2026

RelayCaching: Accelerating LLM Collaboration via Decoding KV Cache Reuse

RelayCaching optimizes multi-agent LLM systems by reusing KV cache across agents to reduce redundant prefill computation and memory usage.

Ax Xianxun Zhu, Zezhong Sun, Imad Rida, Erik Cambria, Junqi Su, Rui Wang, Hui Chen 3/17/2026

FedUAF: Uncertainty-Aware Fusion with Reliability-Guided Aggregation for Multimodal Federated Sentiment Analysis

FedUAF: federated learning framework for multimodal sentiment analysis handling missing modalities and heterogeneous data distributions. arXiv paper.

Ax Ming Wen, Kun Yang, Xin Chen, Jingyu Zhang, Dingding Han, Shiwen Cui, Yuedong Xu 3/17/2026

Pragma-VL: Towards a Pragmatic Arbitration of Safety and Helpfulness in MLLMs

Pragma-VL: method for balancing safety and helpfulness in multimodal LLMs through pragmatic arbitration, addressing jailbreaking and harmful content. arXiv paper.

Ax Rodrigo Tertulino, La\'ercio Alencar 3/17/2026

A Robust Framework for Secure Cardiovascular Risk Prediction: An Architectural Case Study of Differentially Private Federated Learning

FedCVR: federated learning framework for cardiovascular risk prediction using differential privacy. Privacy-preserving ML architecture case study. arXiv paper.

Ax Reva Schwartz, Gabriella Waters 3/17/2026

Real-World AI Evaluation: How FRAME Generates Systematic Evidence to Resolve the Decision-Maker's Dilemma

FRAME methodology for systematic real-world AI evaluation addressing decision-maker needs. Bridges abstract model capabilities with contextual deployment heterogeneity. arXiv paper.

Ax Xinrun Xu, Pi Bu, Ye Wang, B\"orje F. Karlsson, Ziming Wang, Tengtao Song, Qi Zhu, Jun Song, Shuo Zhang, Zhiming Ding, Bo Zheng 3/17/2026

ICPRL: Acquiring Physical Intuition from Interactive Control

ICPRL framework enabling VLMs to acquire physical intuition through interactive pixel-based reinforcement learning in dynamic environments. arXiv paper.

Ax Yuzhang Xie, Yuhua Wu, Ruiyu Wang, Fadi Nahab, Xiao Hu, Carl Yang 3/17/2026

Enhanced Atrial Fibrillation Prediction in ESUS Patients with Hypergraph-based Pre-training

Hypergraph-based pre-training method for atrial fibrillation prediction in stroke patients. Machine learning for healthcare with limited data. arXiv paper.

Ax Henan Wang, Shengwu Xiong, Yifang Zhang, Wenjie Yin, Chen Zhou, Yuqiang Zhang, Pengfei Duan 3/17/2026

FusionCast: Enhancing Precipitation Nowcasting with Asymmetric Cross-Modal Fusion and Future Radar Priors

FusionCast framework for precipitation nowcasting using asymmetric cross-modal fusion of multimodal data. Deep learning optimization for weather prediction. arXiv paper.

Ax Nirmalendu Prakash, Narmeen Oozeer, Michael Lan, Luka Samkharadze, Phillip Howard, Roy Ka-Wei Lee, Dhruv Nathawani, Shivam Raval, Amirali Abdullah 3/17/2026

DreamReader: An Interpretability Toolkit for Text-to-Image Models

DreamReader: unified interpretability toolkit for text-to-image diffusion models enabling activation extraction, causal patching, ablations, and steering. arXiv research.

Ax Mingyu Kim, Young-Heon Kim, Mijung Park 3/17/2026

Safety-Guided Flow (SGF): A Unified Framework for Negative Guidance in Safe Generation

Research paper on safety mechanisms for diffusion and flow models, combining control barrier functions with negative guidance for constrained generation.

Ax Varun Kotte 3/17/2026

Not All Queries Need Rewriting: When Prompt-Only LLM Refinement Helps and Hurts Dense Retrieval

Empirical study of prompt-only LLM query rewriting in RAG pipelines showing domain-dependent effects on dense retrieval.

Ax Viet-Thanh Pham, Lizhen Qu, Zhuang Li, Gholamreza Haffari 3/17/2026

Evidence-based Distributional Alignment for Large Language Models

Evidence-based approach for LLMs to predict population response distributions with stability under domain shift.

Ax Shreyas Bhat Brahmavar, Qiyang Liu, Yang Li, Junier Oliva 3/17/2026

Task Expansion and Cross Refinement for Open-World Conditional Modeling

Semi-supervised framework enabling single model to handle arbitrary conditional queries across heterogeneous datasets.

Ax Vaibhav Mishra 3/17/2026

Preventing Curriculum Collapse in Self-Evolving Reasoning Systems

Self-evolving reasoning system for LLMs that prevents curriculum collapse through balanced problem space exploration.

Ax Quan Liu, Han Zhou, Wenquan Wu, Hua Wu, Sen Su 3/17/2026

Residual Stream Analysis of Overfitting And Structural Disruptions

Analysis of LLM overfitting to safety data via residual stream examination, quantifying false refusal rates in fine-tuned models.

Ax Zixun Luo, Yuhang Fan, Hengyu Lin, Yufei Li, Youzhi Zhang 3/17/2026

Auditing Cascading Risks in Multi-Agent Systems via Semantic-Geometric Co-evolution

Framework for detecting cascading risks in LLM-based multi-agent systems through semantic-geometric analysis of interaction dynamics.

Ax Yeji Kim, Housam Khalifa Bashier Babiker, Mi-Young Kim, Randy Goebel 3/17/2026

Feature-level Interaction Explanations in Multimodal Transformers

Explainability method for multimodal transformers identifying synergistic and redundant cross-modal feature interactions.

Ax Libang Zhao, Qixin Zeng, Hongyin Zhang, Donglin Wang 3/17/2026

Information-Theoretic Constraints for Continual Vision-Language-Action Alignment

Information-theoretic approach to prevent catastrophic forgetting in continual vision-language-action model adaptation for robotics.

Ax Ahmed Elmahdi 3/17/2026

AdaBox: Adaptive Density-Based Box Clustering with Parameter Generalization

AdaBox: grid-based clustering algorithm with adaptive hyperparameters generalizing across datasets without re-optimization.

Ax Zhenyu Zhang, Yixiong Zou, Yuhua Li, Ruixuan Li, Guangyao Chen 3/17/2026

Mind the Discriminability Trap in Source-Free Cross-domain Few-shot Learning

Vision-language model fine-tuning for few-shot cross-domain learning reveals visual discriminability paradox in VLM-based tasks.

Ax Linh-Tam Tran, Sung-Ho Bae 3/17/2026

Post Training Quantization for Efficient Dataset Condensation

Post-training quantization applied to dataset condensation for reducing storage while maintaining model performance.

Ax Daniel Nobrega Medeiros 3/17/2026

PolyGLU: State-Conditional Activation Routing in Transformer Feed-Forward Networks

PolyGLU: transformer activation function enabling dynamic routing among multiple activation functions in feed-forward networks.

Ax Fan Wu, Tsai-Ning Wang, Nicolas Zumarraga, Ning Wang, Markus Kreft, Kevin O'Sullivan, Elgar Fleisch, Oliver Aalami, Paul Schmiedmayer, Robert Jakob, Patrick Langer 3/17/2026

Patient-Level Multimodal Question Answering from Multi-Site Auscultation Recordings

Multimodal QA framework aligning auscultation recordings with frozen LLM embeddings via gated cross-attention. Applies audio-language models to physiological signal interpretation.

Ax Ning Liao, Xiaoxing Wang, Xiaohan Qin, Junchi Yan 3/17/2026

FineRMoE: Dimension Expansion for Finer-Grained Expert with Its Upcycling Approach

FineRMoE: Architecture extending fine-grained expert design across intermediate and output dimensions in MoE models. Improves expert specialization beyond single-dimension optimization.

Ax Zhongxing Xu, Zhonghua Wang, Zhe Qian, Dachuan Shi, Feilong Tang, Ming Hu, Shiyan Su, Xiaocheng Zou, Wei Feng, Dwarikanath Mahapatra, Yifan Peng, Mingquan Lin, Zongyuan Ge 3/17/2026

Thinking in Uncertainty: Mitigating Hallucinations in MLRMs with Latent Entropy-Aware Decoding

Decoding method using latent entropy to mitigate hallucinations in multimodal reasoning models. Extracts contextual uncertainty from token probability distributions.

Ax Sangyoon Lee, Francesca Branzoli, Ma{\l}gorzata Marja\'nska, Patrick Bolan 3/17/2026

Agentic LLM Workflow for MR Spectroscopy Volume-of-Interest Placements in Brain Tumors

Agentic LLM workflow for positioning MR spectroscopy volume-of-interest in brain tumors. Tunes placement for clinician preference and case-specific anatomy.

Ax Jacopo Tagliabue 3/17/2026

Querying Everything Everywhere All at Once: Supervaluationism for the Agentic Lakehouse

System for multi-version agentic analytics on lakehouses. Answers queries across competing data branches using supervaluationary semantics.

Ax Renwei Meng, Haoyi Wu, Jingming Wang, Haoyan Bai 3/17/2026

VulnAgent-X: A Layered Agentic Framework for Repository-Level Vulnerability Detection

VulnAgentX: Layered agentic framework for repository-level vulnerability detection. Integrates risk screening with multi-stage validation across code structure and context.

Ax Youting Wang, Yuan Tang, Yitian Qian, Chen Zhao 3/17/2026

VisualLeakBench: Auditing the Fragility of Large Vision-Language Models against PII Leakage and Social Engineering

Evaluation suite (VisualLeakBench) auditing large vision-language models for privacy vulnerabilities. Tests robustness against OCR injection and PII leakage attacks.

Ax Akshey Sigdel, Rista Baral 3/17/2026

Schema First Tool APIs for LLM Agents: A Controlled Study of Tool Misuse, Recovery, and Budgeted Performance

Experimental study comparing schema-based vs. free-form tool interfaces for LLM agents. Tests JSON Schema specifications with validation diagnostics for improved reliability.

Ax Liao Zhang, Tong Chen, Xiwei Wu, Qi Liu, Xiyu Zhai, Xinqi Wang, Qinxiang Cao 3/17/2026

Neuro-Symbolic Generation and Validation of Memory-Aware Formal Function Specifications

Neuro-symbolic approach for LLM-generated formal function specifications in memory-manipulating code to enable scalable verification.

Ax Vasundra Srinivasan 3/17/2026

Bridging Protocol and Production: Design Patterns for Deploying AI Agents with Model Context Protocol

Design patterns for deploying AI agents with Model Context Protocol, identifying gaps in identity propagation, tool budgeting, and error semantics.

Ax Xiaoyun Liu, Divya Saxena, Jiannong Cao, Yuqing Zhao, Yiying Dong, Penghui Ruan 3/17/2026

GPrune-LLM: Generalization-Aware Structured Pruning for Large Language Models

GPrune-LLM structured pruning method using distribution-robust neuron selection to improve cross-task generalization in compressed models.

Ax Xinhai Wang, Shaopeng Fu, Shu Yang, Liangyu Wang, Tianhang Zheng, Di Wang 3/17/2026

Accelerating Suffix Jailbreak attacks with Prefix-Shared KV-cache

PSKV optimization technique accelerating suffix jailbreak attacks via prefix-shared KV-cache for red-teaming LLMs.

Ax Darren Cheng, Wen-Kwang Tsao 3/17/2026

Agent Privilege Separation in OpenClaw: A Structural Defense Against Prompt Injection

Defense against prompt injection attacks on LLM agents using privilege separation and JSON formatting in OpenClaw platform.

Ax Huamin Chen, Xunzhuo Liu, Junchen Jiang, Bowei He, Xue Liu 3/17/2026

Outcome-Aware Tool Selection for Semantic Routers: Latency-Constrained Learning Without LLM Inference

OATS method for low-latency semantic routing in LLM inference gateways, selecting tools via offline embedding interpolation without GPU cost.

Ax Yu Miao, Zequn Yang, Yake Wei, Ziheng Chen, Haotian Ni, Haodong Duan, Kai Chen, Di Hu 3/17/2026

MIBench: Evaluating LMMs on Multimodal Interaction

MIBench comprehensive benchmark for evaluating multimodal language models on various interaction patterns across modalities.

Ax Gangda Deng, Zhaoling Chen, Zhongming Yu, Haoyang Fan, Yuhong Liu, Yuxin Yang, Dhruv Parikh, Rajgopal Kannan, Le Cong, Mengdi Wang, Qian Zhang, Viktor Prasanna, Xiangru Tang, Xingyao Wang 3/17/2026

EvoClaw: Evaluating AI Agents on Continuous Software Evolution

EvoClaw benchmark evaluating AI agents on continuous software evolution tasks with temporal dependencies and technical debt.

Ax Noam Levy 3/17/2026

Dynamic Sparse Attention: Access Patterns and Architecture

Study of dynamic sparse attention system-level challenges including cache locality, KV fragmentation, and decode throughput optimization.

Ax Sehun Jung, HyunJee Song, Dong-Hee Kim, Reuben Tan, Jianfeng Gao, Yong Jae Lee, Donghyun Kim 3/17/2026

Spatially Grounded Long-Horizon Task Planning in the Wild

Benchmark for assessing whether vision-language models can generate spatially executable robot manipulation plans from task instructions.

Ax Wei Zhuo, Siqiang Luo 3/17/2026

Modality-free Graph In-context Alignment

In-context learning method for graph foundation models enabling cross-domain alignment without modality-specific encoders.

Ax Xin Guan, Yunshan Li, Ze Wang 3/17/2026

NormCode Canvas: Making LLM Agentic Workflows Development Sustainable via Case-Based Reasoning

NormCode Canvas system for LLM agentic workflows using case-based reasoning with compiler-verified scope rules to enable reliable multi-step orchestration.

Ax Ahsan Adeel, M. Bilal 3/17/2026

Scalable Machines with Intrinsic Higher Mental-State Dynamics

Theoretical work on implementing mental-state dynamics in Transformers using biophysical principles and triadic modulation loops for attention mechanisms.

Ax Guanyu Chen, Ruichen Wang, Tianren Zhang, Feng Chen 3/17/2026

Reconciling In-Context and In-Weight Learning via Dual Representation Space Encoding

Framework reconciling in-context and in-weight learning in Transformers through dual representation space encoding.

Ax Jianwei Li, Jung-Eun Kim 3/17/2026

Purifying Generative LLMs from Backdoors without Prior Knowledge or Clean Reference

Backdoor removal method for generative LLMs without prior knowledge of triggers or clean reference models, effective for varied tasks.

Ax Tasos Papazafeiropoulos, Nikolaos Ioannis Bountos, Nikolas Papadopoulos, Ioannis Papoutsis 3/17/2026

Hide and Seek: Investigating Redundancy in Earth Observation Imagery

Investigation of multidimensional redundancy (spectral, temporal, spatial, semantic) in Earth observation data properties.