Ax Anushree Berlia 5/12/2026

Fashion Florence: Fine-Tuning Florence-2 for Structured Fashion Attribute Extraction

Fine-tuned Florence-2 vision-language model using LoRA for extracting structured fashion attributes from clothing images as JSON.

Ax Cheol Young Park, Shou Matsumoto 5/12/2026

Free Energy Manifold: Score-Based Inference for Hybrid Bayesian Networks

Score-based conditional energy model for inference in hybrid Bayesian networks with discrete and continuous variables.

Ax Wenhao Liang, Lin Yue, Wei Emma Zhang, Miao Xu, Mingyu Guo, Olaf Maennel, Weitong Chen 5/12/2026

Probing Routing-Conditional Calibration in Attention-Residual Transformers

Study of routing traces in Attention-Residual transformers for post-hoc calibration, examining whether routing signals improve uncertainty estimates.

Ax Jingchuan Xiao, Xinyi Sui, Cihan Ruan 5/12/2026

Flag Varieties: A Geometric Framework for Deep Network Alignment

Unified geometric framework using flag varieties to theoretically explain alignment phenomena in deep networks including gradient flow and neural collapse.

Ax Danhui Zhang, Zhe Wang, Qing Qing, Jiarui Liu, Wentao Gao, Ziqi Xu, Mingliang Hou, Xikun Zhang, Renqiang Luo 5/12/2026

UFO: A Unified Flow-Oriented Framework for Robust Continual Graph Learning

UFO framework for robust continual graph learning addressing noisy annotations in evolving graphs through flow-oriented methods.

Ax Chunxiao Wang 5/12/2026

Nautilus Compass: Black-box Persona Drift Detection for Production LLM Agents

Nautilus Compass: black-box persona drift detector for production LLM coding agents that tracks constraint violations and memory degradation over long sessions.

Ax Emile Anand, Abdullah Ateyeh, Xinyuan Cao, Max Dabagia 5/12/2026

Continuous Latent Contexts Enable Efficient Online Learning in Transformers

Method enabling efficient online learning in transformers through continuous latent contexts for multi-turn adaptation and feedback-driven decision making.

Ax Tsuyoshi Okita 5/12/2026

Intervention-Based Time Series Causal Discovery via Simulator-Generated Interventional Distributions

SVAR-FM framework for time series causal discovery using physics-based simulators to generate interventional data with flow matching.

Ax Ziyang Wang, Yue Zhang, Shoubin Yu, Ce Zhang, Zengqi Zhao, Jaehong Yoon, Hyunji Lee, Gedas Bertasius, Mohit Bansal 5/12/2026

EgoMemReason: A Memory-Driven Reasoning Benchmark for Long-Horizon Egocentric Video Understanding

EgoMemReason benchmark for evaluating memory and reasoning in ultra-long egocentric video understanding for embodied agents and smart glasses.

Ax Daniel Goldstein, Eugene Cheah 5/12/2026

Key-Value Means

Key-Value Means (KVM): O(N) attention mechanism supporting fixed or growing state for long-context transformers with subquadratic prefill performance.

Ax Xia Hu, Zhenrui Yue, Brian Potetz, Howard Zhou, Leonidas Guibas, Chun-Ta Lu, Zhicheng Wang 5/12/2026

The Cartesian Shortcut: Re-evaluate Vision Reasoning in Polar Coordinate Space

Identifies 'Cartesian Shortcut' vulnerability in vision reasoning benchmarks where MLLMs exploit grid-based layouts through explicit coordinate discretization.

Ax Eslam Zaher, Maciej Trzaskowski, Quan Nguyen, Fred Roosta 5/12/2026

The Geometric Wall: Manifold Structure Predicts Layerwise Sparse Autoencoder Scaling Laws

Theoretical analysis explaining why sparse autoencoders (SAEs) show layerwise scaling variations through manifold geometry, advancing understanding of activation space structure.

Ax Sushrita Rakshit, Hanwen Zhang, Hua Shen 5/12/2026

Pseudo-Deliberation in Language Models: When Reasoning Fails to Align Values and Actions

VALDI benchmark demonstrating 'Pseudo-Deliberation' failure mode where LLMs show reasoning without behavioral alignment, revealing value-action gaps.

Ax Rong Shan, Te Gao, Hang Zheng, Yunjia Xi, Jiachen Zhu, Zeyu Zheng, Yong Yu, Weinan Zhang, Jianghao Lin 5/12/2026

Position: Academic Conferences are Potentially Facing Denominator Gaming Caused by Fully Automated Scientific Agents

Position paper identifying 'Agentic Denominator Gaming' threat where malicious actors deploy AI agents to flood academic conferences with low-quality papers exploiting stable acceptance rates.

Ax Yihang Zhang, Zimeng Huang, Ren Zhai, Yipeng Kang, Tonghan Wang 5/12/2026

NaiAD: Initiate Data-Driven Research for LLM Advertising

NaiAD dataset of 58,999 ad-embedded LLM responses with evaluation metrics for studying LLM-native advertising balancing user experience and platform revenue.

Ax Xuexiang Wen, Hang Yu, Linchao Zhu, Gaoang Wang 5/12/2026

Verifier-Free RL for LLMs via Intrinsic Gradient-Norm Reward

VIGOR: reward model for LLM reinforcement learning that eliminates need for external verifiers by using intrinsic gradient-norm signals, improving scalability to new tasks.

Ax Wu Li, Yigeng Zhou, Zesheng Shi, Yequan Wang, Min Zhang, Jing Li 5/12/2026

Team-Based Self-Play With Dual Adaptive Weighting for Fine-Tuning LLMs

Self-training method using team-based self-play with dual adaptive weighting to improve LLM alignment while reducing dependency on human-labeled data and addressing synthetic data quality issues.

Ax Luan Zhang, Dandan Song, Zhijing Wu, Zhengyu Chen, Chen Zhang, Yuhang Tian, Huipeng Ma, Chenhao Li, Changzhi Zhou, Xudong Li, Shuhao Zhang 5/12/2026

PruneTIR: Inference-Time Tool Call Pruning for Effective yet Efficient Tool-Integrated Reasoning

Inference-time pruning method reducing unnecessary tool calls in LLM reasoning systems while maintaining performance.

Ax Xurui Li, Zhiguo Gan, Jiaming Zhang, Zheng Liu, Diannan Lu 5/12/2026

Novel GPU Boruta algorithms for feature selection from high-dimensional data

GPU-accelerated Boruta feature selection algorithms for high-dimensional datasets with improved computational efficiency.

Ax Chengsong Huang, Haolin Liu, Tong Zheng, Runpeng Dai, Langlin Huang, Jinyuan Li, Zongxia Li, Zhepei Wei, Yu Meng, Jiaxin Huang 5/12/2026

G-Zero: Self-Play for Open-Ended Generation from Zero Data

Self-improvement framework for LLMs using intrinsic rewards without verifiers, enabling autonomous evolution on open-ended tasks.

Ax Urchade Zaratiana, Ash Lewis, George Hurn-Maloney 5/12/2026

GLiNER2-PII: A Multilingual Model for Personally Identifiable Information Extraction

Small 0.3B multilingual NER model for detecting 42 PII entity types across languages and document formats.

Ax Do\u{g}a\c{c} Eldenk, Payal Mohapatra, Yigitcan Comlek, Kaan Oktay, Hongyang Zhang, Stephen Xia 5/12/2026

Attention Drift: What Autoregressive Speculative Decoding Models Learn

Analysis of attention drift phenomenon in speculative decoding drafters showing attention shifts from prompt to generated tokens over speculation chains.

Ax Seth Karten, Joel Zhang, Tersoo Upaa Jr, Ruirong Feng, Wenzhe Li, Chengshuai Shi, Chi Jin, Kiran Vodrahalli 5/12/2026

Continual Harness: Online Adaptation for Self-Improving Foundation Agents

Continual Harness framework for online adaptation of embodied agents with iterative refinement, demonstrated on Pokemon gameplay.

Ax Wolfgang Grieskamp, Teng Zhang, Vineeth Kashyap 5/12/2026

Combining Mechanical and Agentic Specification Inference for Move

Specification inference tool for Move Prover combining weakest-precondition analysis with agentic Claude coding for reduced boilerplate.

Ax Yuna Haseyama, Tomoki Ito, Hiroki Sakaji, Itsuki Noda 5/12/2026

Medical Incident Causal Factors and Preventive Measures Generation Using Tag-based Example Selection in Few-shot Learning

Tag-based few-shot example selection method for prompting LLMs to generate causal factors and preventive measures from medical incident reports.

Ax Terumi Chiba, Yang Luo, Ziyun Cui, Yongsheng Tong, Chao Zhang 5/12/2026

Speech-based Psychological Crisis Assessment using LLMs

LLM-based framework for automated psychological crisis assessment from speech for mental health hotline support.

Ax Xijun Wang, Zhaoyang Liu, Chenyuan Feng, Xiang Chen, Howard H. Yang, Tony Q. S. Quek 5/12/2026

Bridging the Cognitive Gap: A Unified Memory Paradigm for 6G Agentic AI-RAN

Vision for unified memory paradigm for agentic AI in 6G radio access networks to bridge semantic bottleneck in disaggregated architectures.

Ax Xilai Ma, Liye Zhao, Weijun Yao, Haibing Di, Wenya Wang, Jing Li 5/12/2026

Personalizing LLMs with Binary Feedback: A Preference-Corrected Optimization Framework

C-BPO framework for personalizing LLMs using binary preference feedback with inter-user calibration.

Ax Puming Jiang, Meiyi Wang, Kelvin Lin, Ce Hao, Harold Soh 5/12/2026

Guided Streaming Stochastic Interpolant Policy

Guidance mechanism for stochastic interpolant robot policies enabling test-time steering without retraining for dynamic objectives.

Ax Xinyu Zhang, Zhicheng Dou, Deyang Li, Jianjun Tao, Shuo Cheng, Ruifeng Shi, Fangchao Liu, Enrui Hu, Yangkai Ding, Hongbo Wang, Qi Ye, Xuefeng Jin, Zhangchun Zhao 5/12/2026

Swarm Skills: A Portable, Self-Evolving Multi-Agent System Specification for Coordination Engineering

Portable specification for multi-agent coordination as self-improving skills distributed across agent frameworks.

Ax Hyundong Jin, Yo-Sub Han 5/12/2026

NCO: A Versatile Plug-in for Handling Negative Constraints in Decoding

NCO plugin for constraining LLM decoding to prevent undesirable outputs like profanity and PII during generation without post-processing.

Ax Huilin Zhou, Jian Zhao, Yilu Zhong, Zhen Liang, Xiuyuan Chen, Yuchen Yuan, Tianle Zhang, Chi Zhang, Lan Zhang, Xuelong Li 5/12/2026

Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization

Metis framework reformulates LLM red teaming as policy optimization using adversarial MDPs for improved jailbreak discovery.

Ax Jianchao Zhao, Huoren Yang, Hu Yusong, Yuyang Gao, Qiguan Ou, Cong Wan, SongLin Dong, Zhiheng Ma, Yihong Gong 5/12/2026

Retrieve-then-Steer: Online Success Memory for Test-Time Adaptation of Generative VLAs

Test-time adaptation framework for vision-language-action robotic models that uses successful execution history to improve closed-loop reliability.

Ax Tingshu Mou, Jiabo He, Renying Wang, Ce Liu, Hao Yang, Tiehua Zhang, Jingjing Chen, Xingjun Ma 5/12/2026

ViSRA: A Video-based Spatial Reasoning Agent for Multi-modal Large Language Models

ViSRA framework for probing spatial reasoning in multi-modal LLMs through video-based agent without requiring model retraining.

Ax Donghyun Kim, Jaehyoung Park 5/12/2026

Think as Needed: Geometry-Driven Adaptive Perception for Autonomous Driving

Adaptive perception system for autonomous driving that dynamically allocates computation based on scene complexity.

Ax Manyu Li, Ruian He, Chenxi Ma, Weimin Tan, Bo Yan 5/12/2026

MicroWorld: Empowering Multimodal Large Language Models to Bridge the Microscopic Domain Gap with Multimodal Attribute Graph

MicroWorld framework enhances MLLMs for microscopy via multimodal attribute graphs to bridge domain gap with limited training data.

Ax Mateusz Cedro, Marcin Chlebus 5/12/2026

Scaling Vision Models Does Not Consistently Improve Localisation-Based Explanation Quality

Study investigating whether scaling vision models improves localization-based explanation quality across ResNet, DenseNet, ViT architectures.

Ax Sandeep Kumar, Yash Kamdar, Abid Hossain, Bharti Kumari, Tanik Saikh, Asif Ekbal 5/12/2026

When Reviews Disagree: Fine-Grained Contradiction Analysis in Scientific Peer Reviews

NLP method for detecting and analyzing contradictions in scientific peer reviews using fine-grained analysis beyond sentence pairs.

Ax Farzad Nourmohammadzadeh Motlagh, Mehrdad Hajizadeh, Mehryar Majd, Pejman Najafi, Feng Cheng, Christoph Meinel 5/12/2026

When Prompts Become Payloads: A Framework for Mitigating SQL Injection Attacks in Large Language Model-Driven Applications

Security framework addressing SQL injection vulnerabilities in LLM-driven database interfaces through prompt-to-SQL translation.

Ax Guangli Chen, Dianzhao Li, Wenjian Zhong, Bangquan Xie, Ostap Okhrin 5/12/2026

MTA-RL: Robust Urban Driving via Multi-modal Transformer-based 3D Affordances and Reinforcement Learning

MTA-RL framework combines transformers and reinforcement learning for autonomous driving with 3D scene understanding.

Ax Mengzhou Gao, Kaiwei Wang, Pengfei Jiao 5/12/2026

One-Step Graph-Structured Neural Flows for Irregular Multivariate Time Series Classification

Neural flows method for irregular multivariate time series classification that models inter-variable interactions in one-step mapping.

Ax Jihyeon Baek, Seunghoon Lee, Gitaek Kwon, Doohyun Park 5/12/2026

A Comparative Study of Machine Learning and Deep Learning for Out-of-Distribution Detection

Comparative study of ML vs DL for out-of-distribution detection in medical imaging, showing ML can match DL performance on constrained datasets.

Ax Sijia Chen, Hang Yin, Shunfan Zhou 5/12/2026

LegalCiteBench: Evaluating Citation Reliability in Legal Language Models

LegalCiteBench evaluates citation reliability in legal LLMs, testing case authority provision to address hallucination and fabrication risks.

Ax Yulin Zhang, He Cao, Zihao Jiang, Chenyi Zi, Zhipeng Zhou, Zijing Liu, Yu Li, Jia Li, Ziqi Gao 5/12/2026

ProteinOPD: Towards Effective and Efficient Preference Alignment for Protein Design

Protein design method combining protein language models with preference alignment to avoid catastrophic forgetting while steering toward desired functions.

Ax Nicola Novello, Andrea M. Tonello 5/12/2026

Empty SPACE: Cross-Attention Sparsity for Concept Erasure in Diffusion Models

Concept erasure in diffusion models via cross-attention sparsity maintaining effectiveness when scaling to larger architectures like SDXL.

Ax Maik Larooij, David Graus 5/12/2026

To Redact, or not to Redact? A Local LLM Approach to Deliberative Process Privilege Classification

Local LLM approach for automated deliberative process privilege classification in FOIA document redaction without external APIs.

Ax Peiru Yang, Haoran Zheng, Tong Ju, Shiting Wang, Wanchun Ni, Jiajun Liu, Shangguang Wang, Yongfeng Huang, Tao Qi 5/12/2026

Knowledge Poisoning Attacks on Medical Multi-Modal Retrieval-Augmented Generation

Security analysis of knowledge poisoning attacks on medical multimodal RAG systems, demonstrating adversarial risks in retrieval databases.

Ax Baibei Ji, Xiaoyang Weng, Juntao Li, Zecheng Tang, Yihang Lou, Min Zhang 5/12/2026

MemReread: Enhancing Agentic Long-Context Reasoning via Memory-Guided Rereading

MemReread improves agentic long-context reasoning by memory-guided rereading to recover latent evidence when processing document chunks linearly.

Ax Haaris Mehmood, Jie Xu, Karthikeyan Saravanan, Rogier Van Dalen, Mete Ozay 5/12/2026

DP-LAC: Lightweight Adaptive Clipping for Differentially Private Federated Fine-tuning of Language Models

DP-LAC enables differentially private federated fine-tuning of LLMs with lightweight adaptive gradient clipping for privacy-preserving training.

Ax Anton Bazdyrev, Ivan Bashtovyi, Ivan Havlytskyi, Oleksandr Kharytonov, Artur Khodakovskyi 5/12/2026

Qwen Goes Brrr: Off-the-Shelf RAG for Ukrainian Multi-Domain Document Understanding

RAG system using Qwen for Ukrainian multi-domain document understanding with PDF chunking, dense retrieval, and answer reranking.