Ax Nimrod Shabtay, Moshe Kimhi, Artem Spector, Sivan Haray, Ehud Rivlin, Chaim Baskin, Raja Giryes, Eli Schwartz 3/19/2026

Look Where It Matters: High-Resolution Crops Retrieval for Efficient VLMs

AwaRes framework for efficient vision-language models using spatial-on-demand high-resolution crop retrieval to balance accuracy and computational cost.

Ax Abderrahmene Boudiaf, Irfan Hussain, Sajid Javed 3/19/2026

AgriChat: A Multimodal Large Language Model for Agriculture Image Understanding

Multimodal LLM for agriculture with Vision-to-Verified-Knowledge pipeline addressing lack of domain-specific agricultural datasets and expertise.

Ax Adam Massimo Mazzocchetti 3/19/2026

Cryptographic Runtime Governance for Autonomous AI Systems: The Aegis Architecture for Verifiable Policy Enforcement

Aegis architecture for cryptographic runtime governance of autonomous AI systems, enforcing policy constraints as verifiable execution conditions.

Ax Wen Yan, Qianye Yang, Yipei Wang, Shonit Punwani, Mark Emberton, Vasilis Stavrinides, Yipeng Hu, Dean Barratt 3/19/2026

On the Degrees of Freedom of Gridded Control Points in Learning-Based Medical Image Registration

GridReg framework investigating optimal sparse control-point parameterization for learning-based medical image registration.

Ax Yujia Yang, Yuanxiang Wang, Zhenyu Guan, Tiankun Yang, Chenxi Bao, Haopeng Jin, Jinwen Luo, Xinyu Zuo, Lisheng Duan, Haijin Liang, Jin Ma, Xinming Wang, Ruiwen Tao, Hongzhu Yi 3/19/2026

Omni IIE Bench: Benchmarking the Practical Capabilities of Image Editing Models

Omni IIE Bench, a high-quality benchmark for evaluating instruction-based image editing models across tasks of varying semantic scales.

Ax Ke Wang, Yanfei Cao, Xiangzhi Tao, Naijie Gu, Jun Yu, Zhengdong Wang, Shouyang Dong, Fan Yu, Cong Wang, Yang Luo 3/19/2026

Joint Optimization of Storage and Loading for High-Performance 3D Point Cloud Data Processing

Optimization techniques for storage and loading of 3D point cloud data in computer vision and deep learning applications.

Ax Kun Luo, Xiaoguang Ma 3/19/2026

EmergeNav: Structured Embodied Inference for Zero-Shot Vision-and-Language Navigation in Continuous Environments

EmergeNav framework for zero-shot vision-and-language navigation using structured embodied inference to improve VLM execution in continuous environments.

Ax Utkarsh Grover (University of South Florida, Tampa, USA), Ravi Ranjan (Florida International University, Miami, USA), Mingyang Mao (University of South Florida, Tampa, USA), Trung Tien Dong (University of South Florida, Tampa, USA), Satvik Praveen (University of South Florida, Tampa, USA), Zhenqi Wu (University of South Florida, Tampa, USA), J. Morris Chang (University of South Florida, Tampa, USA), Tinoosh Mohsenin (Johns Hopkins University, Baltimore, USA), Yi Sheng (University of South Florida, Tampa, USA), Agoritsa Polyzou (Florida International University, Miami, USA), Eiman Kanjo (Nottingham Trent University, Nottingham, United Kingdom, Imperial College London, London, United Kingdom), Xiaomin Lin (University of South Florida, Tampa, USA) 3/19/2026

Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies

Survey of deployment constraints and mitigation strategies for foundation models in edge embodied systems with real-time control requirements.

Ax Hisayuki Yokomizo, Taiki Miyanishi, Yan Gang, Shuhei Kurita, Nakamasa Inoue, Yusuke Iwasawa 3/19/2026

PhysQuantAgent: An Inference Pipeline of Mass Estimation for Vision-Language Models

PhysQuantAgent inference pipeline for mass estimation in vision-language models to improve robotic manipulation by enabling physical property reasoning.

Ax Wenqiang Huang, Susu Fang, Xuhang Gu, Shen'ao Xue, Huanhuan Xing, Junjie Jiang, Junying Zhang, Shen Zhou, Zheng Luo, Jin Zhang, Fangping Ouyang, Shanshan Wang 3/19/2026

Machine intelligence supports the full chain of 2D dendrite synthesis

Machine learning framework for optimizing 2D dendrite synthesis via active learning, synthesis automation, and mechanism discovery.

Ax Alejandro Paredes La Torre 3/19/2026

Adversarial attacks against Modern Vision-Language Models

Adversarial robustness evaluation of open-source VLMs (LLaVA, Qwen2.5-VL) against gradient-based attacks in e-commerce environments.

Ax Liangbin Huang, Xiaohua Liao, Chaoqun Cui, Shijing Wang, Zhaolong Huang, Yanlong Du, Wenji Mao 3/19/2026

CineSRD: Leveraging Visual, Acoustic, and Linguistic Cues for Open-World Visual Media Speaker Diarization

CineSRD system for speaker diarization in films and TV series using visual, acoustic, and linguistic cues in open-world scenarios.

Ax Zhaoyuan Qiu, Ken Chen, Xiangwei Wang, Yu Xia, Sachith Seneviratne, Saman Halgamuge 3/19/2026

MSRAMIE: Multimodal Structured Reasoning Agent for Multi-instruction Image Editing

MSRAMIE training-free agent framework for multi-instruction image editing using multimodal foundation models without additional dataset annotation.

Ax Trung V. Phan, Tri Gia Nguyen, Thomas Bauschert 3/19/2026

DeepStage: Learning Autonomous Defense Policies Against Multi-Stage APT Campaigns

DeepStage framework using deep reinforcement learning and graph neural networks for autonomous defense against multi-stage APT campaigns.

Ax Wonseon Lim, Hyejeong Im, Dae-Won Kim 3/19/2026

Continual Multimodal Egocentric Activity Recognition via Modality-Aware Novel Detection

Method for continuous learning in multimodal egocentric activity recognition using modality-aware novelty detection in non-stationary streams.

Ax Pengyu Zhang, Klim Zaporojets, Jie Liu, Jia-Hong Huang, Paul Groth 3/19/2026

Are a Thousand Words Better Than a Single Picture? Beyond Images -- A Framework for Multi-Modal Knowledge Graph Dataset Enrichment

Beyond Images pipeline for automatic enrichment of multi-modal knowledge graphs by retrieving and converting images to text descriptions.

Ax Vincent Gurgul, Robin Gubela, Stefan Lessmann 3/19/2026

The State of Generative AI in Software Development: Insights from Literature and a Developer Survey

Survey of 65 developers examining generative AI impact on software development lifecycle tasks, showing 70% reduction in time for design, implementation, testing, documentation.

Ax Sa Xiao, Hao Jing, Honglu Sun, Haoyu Li 3/19/2026

Implementation of tangent linear and adjoint models for neural networks based on a compiler library tool

TorchNWP compiler library tool for coupling AI components with numerical models, improving cross-language compatibility and data transfer efficiency.

Ax Jiabo Huang, Zhizhong Li, Sina Sajadmanesh, Weiming Zhuang, Lingjuan Lyu 3/19/2026

Empirical Recipes for Efficient and Compact Vision-Language Models

Empirical analysis and optimization recipes for efficient deployment of vision-language models in resource-constrained environments, addressing inference bottlenecks.

Ax Lifu Tu, Rongguang Wang, Tao Sheng, Sujjith Ravi, Dan Roth 3/19/2026

LLM NL2SQL Robustness: Surface Noise vs. Linguistic Variation in Traditional and Agentic Settings

Robustness evaluation benchmark for NL2SQL systems under perturbations, comparing traditional and agentic LLM settings in dynamic database environments.

Ax Shenzhi Wang, Shixuan Liu, Jing Zhou, Chang Gao, Xiong-Hui Chen, Binghai Wang, An Yang, Shiji Song, Bowen Yu, Gao Huang, Junyang Lin 3/19/2026

HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning

Data synthesis approach for improving vision-language model reasoning through multi-hop chains, addressing perception and hallucination errors.

Ax Shubham Gupta, Adarsh Arigala, B. R. Dilleswari, Sri Rama Murty Kodukula 3/19/2026

Shared Representation Learning for Reference-Guided Targeted Sound Detection

Method for targeted sound detection using shared representation learning and conditional embedding vectors for audio source separation.

Ax Nazia Riasat 3/19/2026

Dependence Fidelity and Downstream Inference Stability in Generative Models

Proposes dependence fidelity as evaluation criterion for generative models to assess preservation of multivariate dependence structures in synthetic data.

Ax Abinav Rao, Sujan Rachuri 3/19/2026

Do Understanding and Generation Fight? A Diagnostic Study of DPO for Unified Multimodal Models

Systematic study of DPO alignment applied to unified multimodal models, finding generation quality resists alignment across tested conditions.

Ax Wenhao Zhao, Qiran Zou, Rushi Shah, Yudi Wu, Zhouhan Lin, Dianbo Liu 3/19/2026

Early Quantization Shrinks Codebook: A Simple Fix for Diversity-Preserving Tokenization

Study of vector quantization behavior in tokenization for LLMs and generative models, investigating codebook diversity issues and proposing fixes.

Ax Yi Zhou, Basel Shbita 3/19/2026

Evaluating Ill-Defined Tasks in Large Language Models

Research on evaluating LLMs on ill-defined tasks with ambiguous criteria, analyzing why existing benchmarks fail for complex instruction following.

Ax Lucas Bandarkar, Alan Ansell, Trevor Cohn 3/19/2026

Large Reasoning Models Struggle to Transfer Parametric Knowledge Across Scripts

Analysis of large reasoning LLMs showing cross-lingual knowledge transfer gaps are primarily script barriers rather than language or family effects.

Ax Weikun K. Zhang, Rohan Pandey, Bhaumik Mehta, Kaijie Jin, Naomi Morato, Archit Ganapule, Michael Ruofan Zeng, Jarod Alper 3/19/2026

CircuitBuilder: From Polynomials to Circuits via Reinforcement Learning

RL agent trained with AlphaZero-style loop to discover efficient arithmetic circuits for computing polynomials using addition and multiplication gates.

Ax Lucas Bandarkar, Alan Ansell, Trevor Cohn 3/19/2026

Knowledge Localization in Mixture-of-Experts LLMs Using Cross-Lingual Inconsistency

Framework for localizing parametric knowledge in mixture-of-experts LLMs using cross-lingual inconsistency as interpretability signal.

Ax Lu Yan, Xuan Chen, Xiangyu Zhang 3/19/2026

When the Specification Emerges: Benchmarking Faithfulness Loss in Long-Horizon Coding Agents

SLUMP benchmark measures faithfulness loss when coding agent specifications emerge through interaction, tracking design commitments across long coding sessions.

Ax Zacharie Bugaud 3/19/2026

Hidden Clones: Exposing and Fixing Family Bias in Vision-Language Model Ensembles

Study of family bias in Vision-Language Model ensembles, showing models from same family share correlated errors that reduce effective ensemble diversity.

Ax Taiwo Onitiju, Iman Vakilinia 3/19/2026

Security Assessment and Mitigation Strategies for Large Language Models: A Comprehensive Defensive Framework

Comprehensive security assessment comparing vulnerability of major LLM architectures to adversarial attacks, providing defensive framework for secure deployment.

Ax Yasi Zhang, Tianyu Chen, Mingyuan Zhou, Oscar Leong, Ying Nian Wu, Michal Lukasik 3/19/2026

REAL: Regression-Aware Reinforcement Learning for LLM-as-a-Judge

REAL method applies regression-aware RL to train LLMs-as-judges that assign numeric scores, preserving ordinal structure in evaluation tasks.

Ax Shuvendu K. Lahiri 3/19/2026

Intent Formalization: A Grand Challenge for Reliable Coding in the Age of AI Agents

Position paper arguing intent formalization—translating informal natural language requirements to precise specifications—is critical for reliable AI-generated code.

Ax Reshabh K Sharma, Linxi Jiang, Zhiqiang Lin, Shuo Chen 3/19/2026

PAuth - Precise Task-Scoped Authorization For Agents

PAuth authorization framework enables precise task-scoped permissions for AI agents interacting with web services, replacing operator-scoped OAuth models.

Ax Jacob Piland, Byron Dowling, Christopher Sweet, Adam Czajka 3/19/2026

Generalist Multimodal LLMs Gain Biometric Expertise via Human Salience

Study showing generalist multimodal LLMs can gain biometric expertise for iris presentation attack detection via human salience training.

Ax Shenao Yan, Shimaa Ahmed, Shan Jin, Sunpreet S. Arora, Yiwei Cai, Yizhen Wang, Yuan Hong 3/19/2026

Detecting Data Poisoning in Code Generation LLMs via Black-Box, Vulnerability-Oriented Scanning

Black-box scanning approach detects data poisoning attacks in code generation LLMs by checking for vulnerability-inducing code generation patterns.

Ax Patrick Levi 3/19/2026

Towards Unsupervised Adversarial Document Detection in Retrieval Augmented Generation Systems

Unsupervised detection of adversarial documents in retrieval augmented generation systems to defend against context manipulation attacks.

Ax Parsa Mirtaheri, Mikhail Belkin 3/19/2026

Catching rationalization in the act: detecting motivated reasoning before and after CoT via activation probing

Activation probing technique detects motivated reasoning in LLMs where chain-of-thought rationalizes answers influenced by injected hints without acknowledging them.

Ax Haoyang Fang, Shuai Zhang, Yifei Ma, Hengyi Wang, Cuixiong Hu, Katrin Kirchhoff, Bernie Wang, George Karypis 3/19/2026

OPERA: Online Data Pruning for Efficient Retrieval Model Adaptation

OPERA data pruning framework improves efficiency and effectiveness of dense retriever finetuning by exploiting heterogeneous training pair quality.

Ax Eden Saig, Tamar Garbuz, Ariel D. Procaccia, Inbal Talgam-Cohen, Jamie Tucker-Foltz 3/19/2026

Adaptive Contracts for Cost-Effective AI Delegation

Framework for adaptive contracts in AI delegation that selectively performs detailed evaluation to balance noise reduction against evaluation costs.

Ax Federico Albanese, Pablo Ronco, Nicol\'as D'Ippolito 3/19/2026

Anonymous-by-Construction: An LLM-Driven Framework for Privacy-Preserving Text

LLM-driven on-premise pipeline for anonymizing text by replacing PII with realistic type-consistent surrogates while preserving data utility.

Ax Eilam Shapira, Moshe Tennenholtz, Roi Reichart 3/19/2026

Alignment Makes Language Models Normative, Not Descriptive

Empirical study comparing 120 base-aligned LLM pairs on 10K human decisions, showing aligned models underperform at predicting actual human behavior in strategic games.

Ax Prajwal Panth, Agniva Maiti 3/19/2026

TharuChat: Bootstrapping Large Language Models for a Low-Resource Language via Synthetic Data and Human Validation

TharuChat applies synthetic data generation and human validation to bootstrap LLMs for Tharu, a low-resource Indo-Aryan language spoken by 1.7M people.

Ax Boyong Wu, Sanghwan Kim, Zeynep Akata 3/19/2026

From Drop-off to Recovery: A Mechanistic Analysis of Segmentation in MLLMs

Analysis of spatial understanding capabilities in multimodal LLMs for segmentation tasks via layerwise probing and attention mechanisms.

Ax Sohaib Errabii, Olivier Sentieys, Marcello Traiola 3/19/2026

KANtize: Exploring Low-bit Quantization of Kolmogorov-Arnold Networks for Efficient Inference

Research on low-bit quantization techniques for Kolmogorov-Arnold Networks to enable efficient inference.

Ax Jane Wang, Timothy Keyes, April S Liang, Stephen P Ma, Jason Shen, Jerry Liu, Nerissa Ambers, Abby Pandya, Rita Pandya, Jason Hom, Natasha Steele, Jonathan H Chen, Kevin Schulman 3/19/2026

Deployment and Evaluation of an EHR-integrated, Large Language Model-Powered Tool to Triage Surgical Patients

Study deploying LLM-based tool integrated with EHR system to automate surgical patient triage at Stanford Health Care.

Ax Mohamed Mejri, Ashiqur Rasul, Abhijit Chatterjee 3/19/2026

DANCE: Dynamic 3D CNN Pruning: Joint Frame, Channel, and Feature Adaptation for Energy Efficiency on the Edge

DANCE dynamically prunes 3D CNNs at frame, channel, and feature levels to maximize energy efficiency for edge video processing.

Ax Weihua Xiao, Jason Blocklove, Matthew DeLorenzo, Johann Knechtel, Ozgur Sinanoglu, Kanad Basu, Jeyavijayan Rajendran, Siddharth Garg, Ramesh Karri 3/19/2026

GUIDE: GenAI Units In Digital Design Education

GUIDE is open courseware with runnable Colab labs teaching generative AI with standardized units of slides, videos, labs, and papers.

Ax Haiyang Yan, Hongyun Zhou, Peng Xu, Xiaoxue Feng, Mengyi Liu 3/19/2026

Symphony: A Cognitively-Inspired Multi-Agent System for Long-Video Understanding

Multi-agent MLLM system for long-form video understanding using cognitive inspiration and task decomposition with improved context management.