Ax Hao Ren, Zetong Bi, Yiming Zeng, Le Zheng, Zhi Li, Zhaoliang Wan, Lu Qi, Hui Cheng 5/29/2026

Fisher-Preserving Guidance: Training-Free Manifold Constraints for Safe Diffusion Control

Training-free guidance method for diffusion models in visual navigation, constraining predictions to training manifold.

Ax Junhui Ding, Xinchen Zhang, Xiaohui Xie, Shinan Liu 5/29/2026

TraceCodec: A Compiler-Backed Neural Codec for Stateful Multi-Flow Network Traffic Traces

Neural codec for network packet capture synthesis using compiler-backed approach to generate high-fidelity PCAP traces.

Ax Anisha Saha, Varsha Suresh, Teodora Kamova, Sophia Wiedmann, Timothy Hospedales, Vera Demberg 5/29/2026

MuPHI: Learning Implicit Multimodal Harm Reasoning via Semantically Grounded Reward Optimization

Vision-language model evaluation framework for detecting compositional multimodal harm through intent-aware cross-modal reasoning.

Ax Mark Vero, Fabian Kaczmarczyck, Ivan Petrov, Ilia Shumailov, Jamie Hayes, Niels Heinen, Tianqi Fan, Luca Invernizzi, Martin Vechev 5/29/2026

Honeyval: A Comprehensive Evaluation Framework for LLM-powered HTTP Honeypots

Evaluation framework for LLM-powered HTTP honeypots, addressing lack of unified testing methodology for security simulation systems.

Ax Anna Wimbauer, Jonas M\"oller, Erik Imgrund, Konrad Rieck 5/29/2026

Fingerprinting Inference Systems of Large Language Models

Study demonstrating LLM inference system components (engine, hardware, attention backend) create fingerprinting signatures via numerical deviations with security implications.

Ax V\'ictor Gallego 5/29/2026

Discovering Cooperative Pipelines: Autoresearch for Sequential Social Dilemmas

Autoresearch system where AI agent autonomously redesigns LLM policy-synthesis pipelines for multi-agent sequential social dilemmas through code editing and iterative evaluation.

Ax Tanmoy Chakraborty, Ayan Sengupta, Suparna Bhattacharya, Partha Pratim Chakrabarti, Amlan Chakrabarti, Supratik Chakraborty, Partha Pratim Das, Lipika Dey, Richa Singh, Mayank Vatsa 5/29/2026

Latent Performance Profiling of Large Language Models

Framework for profiling LLM latent capabilities beyond benchmark accuracy, addressing contamination and reliability issues in standardized evaluations.

Ax Shutong Ding, Zejia Zhong, Zhongyi Wang, Ke Hu, Bikang Pan, Jingya Wang, Ye Shi 5/29/2026

Sample-Efficient Diffusion-based Reinforcement Learning with Critic Guidance

Reinforcement learning method combining diffusion policies with critic guidance to improve sample efficiency and exploitation of Q-value information.

Ax Zhuguanyu Wu, Ruihao Gong, Yang Yong, Yushi Huang, Xiangyu Fan, Lei Yang, Dahua Lin, Xianglong Liu 5/29/2026

SGMD: Score Gradient Matching Distillation for Few-Step Video Diffusion Distillation

Score gradient matching distillation approach for accelerating few-step video diffusion models while preserving motion dynamics better than reverse-KL matching.

Ax Lixuan Guo, Yifei Wang, Tiansheng Wen, Aosong Feng, Stefanie Jegelka, Chenyu You 5/29/2026

No More K-means:Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval

Sparse coding method for efficient multi-vector retrieval avoiding K-means clustering to reduce storage and computational bottlenecks in token-level dense retrieval.

Ax Selim Kuzucu, Alessio Tonioni, Vasile Lup, Bernt Schiele, Federico Tombari, Muhammad Ferjad Naeem 5/29/2026

PARCEL: Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding

PARCEL method for efficient vision-language model inference through elastic visual-token compression with spatial-temporal conditioning under aggressive compression.

Ax Jingda Wu, Changxiao Cai 5/29/2026

Diffusion Models Are Statistically Optimal for Learning Low-Dimensional Multi-Modal Distributions

Theoretical analysis proving diffusion models are statistically optimal for learning low-dimensional multi-modal distributions under realistic regularity conditions.

Ax Caleb DeLeeuw 5/29/2026

BioRefusalAudit: Auditing Biosecurity Refusal Depth Using General and Domain-Fine-Tuned Sparse Autoencoders

Audit framework using sparse autoencoders to evaluate whether LLM refusals on biosecurity queries are structurally robust or exploitable through prompt variations.

Ax Hanning Lu, Yingguang Yang, Jinwei Su, Yang Liu, Zhaoqian Yao, Yaoming Li, Taoran Liang, Ziyi Zhang, Ran Ran, Kefu Xu, Bin Chong 5/29/2026

SAHG: Sector-Anisotropic Hyperbolic Graph Model for Social Bot Detection

Graph-based model using hyperbolic geometry to detect LLM-driven social bots by exploiting relational patterns and community structures.

Ax Xingzhou Pang, Yifan Hou, Junling Wang, Mrinmaya Sachan 5/29/2026

Unveiling the Visual Counting Bottleneck in Vision-Language Models

Investigation of vision-language models' systematic failures in visual counting by decomposing task into visual individuation, magnitude awareness, and symbolic mapping stages.

Ax Travis Lelle 5/29/2026

Token-Level Generalization in LoRA Adapter Backdoors: Attack Characterization and Behavioral Detection

Security study showing LoRA adapters for LLMs can be backdoored through data poisoning while maintaining baseline performance, with token-level generalization.

Ax Panyu Jiao, Shuizhou Chen, Yiheng Shen, Yuyang Wang, Runhai Ouyang, Wei Xie 5/29/2026

What drives performance in molecular MPNNs? An operator-level factorial benchmark

Factorial benchmark decomposing molecular message-passing neural networks into operator families to identify performance drivers in molecular property prediction.

Ax Haoming Xu, Weihong Xu, Zongrui Li, Mengru Wang, Yunzhi Yao, Chiyu Wu, Jin Shang, Yu Gong, Shumin Deng 5/29/2026

When Should Models Change Their Minds? Contextual Belief Management in Large Language Models

Framework for contextual belief management in LLMs with BeliefTrack benchmark for measuring state update decisions.

Ax Yicheng Tao, Yiqun Wang, Xiangchen Song, Xin Luo, Kai Liu, Jie Liu 5/29/2026

GRASP: Plan-Guided Graph Retrieval with Adaptive Fusion and Reranking on Semi-Structured Knowledge Bases

Retrieval system for semi-structured knowledge bases combining graph traversal with adaptive fusion and reranking.

Ax Ziwen Xu, Haiwen Hong, Linsong Yu, Benglei Cui, Longtao Huang, Hui Xue, Ningyu Zhang 5/29/2026

How LoRA Remembers? A Parametric Memory Law for LLM Finetuning

Quantitative analysis of LoRA memory capacity limits and dynamics for LLM finetuning and knowledge updates.

Ax Anany Kotawala 5/29/2026

Resolution Diagnostics for Paired LLM Evaluation

Statistical analysis revealing unresolved pairwise comparisons in LLM leaderboard rankings due to insufficient sample sizes.

Ax Qinpei Luo, Ruichun Ma, Xinyu Zhang, Lili Qiu 5/29/2026

SchGen: PCB Schematic Generation with Semantic-Grounded Code Representations

LLM for generating PCB schematics from natural language using semantic-grounded code representations.

Ax Yaxin Luo, Jiacheng Cui, Xiaohan Zhao, Xinyi Shang, Jiacheng Liu, Xinyue Bi, Zhaoyi Li, Zhiqiang Shen 5/29/2026

LLMSurgeon: Diagnosing Data Mixture of Large Language Models

Method to audit and estimate pretraining data composition of LLMs from generated text outputs via data mixture surgery.

Ax Jusuk Lee, Seungjae Lee, Jonghun Shin, Hoseong Jung, Sungha Kim, Daesol Cho, H. Jin Kim, Jia-Bin Huang, Furong Huang 5/29/2026

DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation

Multimodal pre-training framework for robot perception integrating visual, language, and dynamics understanding upstream.

Ax Seunghan Lee, Taeyoung Park, Kibok Lee 5/29/2026

Dataset-Driven Channel Masks in Transformers for Multivariate Time Series

Transformer architecture with dataset-driven channel masking for capturing dependencies in multivariate time series.

Ax Quan Zhou, Changhua Pei, Fei Sun, Jing Han, Zhengwei Gao, Dan Pei, Haiming Zhang, Gaogang Xie, Jianhui Li 5/29/2026

KAN-AD: Time Series Anomaly Detection with Kolmogorov-Arnold Networks

Time series anomaly detection using Kolmogorov-Arnold Networks to model normal behavior patterns in monitoring systems.

Ax Kosio Beshkov 5/29/2026

A Quotient Homology Theory of Representation in Neural Networks

Theoretical analysis of ReLU neural network representations using quotient homology and hyperplane arrangements for understanding network geometry.

Ax Petr Sokerin, Maria Kovaleva, Ekaterina Boyarina, Pavel Tikhomirov, Denis Vorobiyov, Alexey Zaytsev 5/29/2026

Looking around you: external information enhances representations for event sequences

Representation learning method for event sequences that incorporates contextual information from co-occurring sequences.

Ax Shutong Ding, Yimiao Zhou, Ke Hu, Xi Yao, Junchi Yan, Xiaoying Tang, Ye Shi 5/29/2026

Diffusion-based learning framework for Constrained Nonconvex Optimization with Weighted Bootstrapped Refinement

Diffusion-based optimization framework for constrained nonconvex problems with constraint enforcement mechanisms.

Ax Seoungbin Bae, Dabeen Lee 5/29/2026

Neural Logistic Bandits

Study of neural logistic bandits problem for learning reward functions using neural networks with improved theoretical dependencies.

Ax Yongding Tian, Zaid Al-Ars, Maksim Kitsak, Peter Hofstee 5/29/2026

Connecting Independently Trained Modes via Layer-Wise Connectivity

Research on constructing low-loss paths between independently trained neural network models through layer-wise connectivity.

Ax Jiaxiang Yi, Miguel A. Bessa 5/29/2026

Cooperative Variance Estimation and Bayesian Neural Networks for Disentangling Aleatoric and Epistemic Uncertainties

Method combining variance estimation networks and Bayesian neural networks to predict both aleatoric and epistemic uncertainty in neural networks.

Ax Dan Qiao, Wenhao Li, Shanchao Yang, Hongyuan Zha, Baoxiang Wang 5/29/2026

Offline Multi-agent Reinforcement Learning via Sequential Score Decomposition

Offline multi-agent RL addressing distributional shifts in joint action spaces using sequential score decomposition for multimodal policies.

Ax Yuheng Lei, Sitong Mao, Shunbo Zhou, Hongyuan Zhang, Xuelong Li, Ping Luo 5/29/2026

Dynamic Mixture of Progressive Parameter-Efficient Expert Library for Lifelong Robot Learning

Lifelong robot learning framework using parameter-efficient expert library with dynamic mixture for forward transfer and catastrophic forgetting mitigation.

Ax Wenhao Dai, Haodong Deng, Mengfei Rong, Xinyu Yang, Hongyu Liu, Fangxin Liu, Hailong Yang, Qianwen Cao, Qingxiao Sun 5/29/2026

Accelerating Sparse Transformer Inference on GPU

GPU optimization techniques for sparse Transformer inference with dynamic operator fusion to accelerate LLM computation.

Ax Yang Dai, Oubo Ma, Longfei Zhang, Xingxing Liang, Xiaochun Cao, Shouling Ji, Jiaheng Zhang, Jincai Huang, Li Shen 5/29/2026

TrojanTO: Action-Level Backdoor Attacks against Trajectory Optimization Models

Backdoor attack method against trajectory optimization models in offline RL targeting action sequences rather than rewards.

Ax Phoomraphee Luenam, Andreas Spanopoulos, Amit Sant, Thomas Hofmann, Sotiris Anagnostidis, Sidak Pal Singh 5/29/2026

Model Fusion via Retrofitting

Neuron-centric model fusion method combining independently trained networks without retraining, handling permutation invariance and non-IID data.

Ax Shibin Su, Guoqiang Liang, De Cheng, Shizhou Zhang, Lingyan Ran 5/29/2026

Multi-level Collaborative Distillation Meets Global Workspace Model: A Unified Framework for OCIL

Online class-incremental learning framework combining collaborative distillation with global workspace model for continuous learning.

Ax Kevin Bachelor, Sanya Murdeshwar, Daniel Sabo, Razvan Marinescu 5/29/2026

Active Learning for Machine Learning Driven Molecular Dynamics

Active learning framework for machine-learned coarse-grained molecular dynamics using RMSD-based frame selection.

Ax Michael Sullivan, Alexander Koller 5/29/2026

GRPO is Secretly a Process Reward Model

Theoretical proof that GRPO RL algorithm with outcome reward models is equivalent to process reward models with Monte-Carlo-based objectives.

Ax Tian Lan, Hao Duong Le, Jinbo Li, Wenjun He, Meng Wang, Chenghao Liu, Chen Zhang 5/29/2026

Towards Foundation Models for Zero-Shot Time Series Anomaly Detection: Leveraging Synthetic Data and Relative Context Discrepancy

Foundation model for time series anomaly detection using synthetic data and relative context discrepancy for zero-shot generalization.

Ax Kosio Beshkov, Anders Malthe-S{\o}renssen 5/29/2026

Towards Understanding the Shape of Representations in Protein Language Models

Interpretability study of protein language models examining how sequences transform to hidden representations and encoded information.

Ax Youngbin Choi, Minjong Lee, Saemi Moon, Seunghyuk Cho, Chaehyeon Chung, MoonJeong Park, Dongwoo Kim 5/29/2026

In-Place Feedback: Reliable Refinement for Multi-Turn Expert-LLM Collaboration

LLM interaction paradigm where users edit model outputs in-place and model continues generation, improving multi-turn feedback integration for error correction.

Ax Feiyang Wu, Ye Zhao, Anqi Wu 5/29/2026

Distributional Inverse Reinforcement Learning

Distributional IRL framework for offline learning that captures reward distributions and expert behavior uncertainty using stochastic dominance.

Ax Xinsong Feng, Leshu Tang, Chenan Wang, Haipeng Chen 5/29/2026

Offline Reinforcement Learning with Generative Trajectory Policies

Offline RL method combining generative models with fast single-step consistency policies to balance speed and performance in trajectory generation.

Ax Coen Adler, Yuxin Chang, Felix Draxler, Samar Abdi, Padhraic Smyth 5/29/2026

Beyond Accuracy: Are Time Series Foundation Models Well-Calibrated?

Study examining calibration properties of time series foundation models beyond accuracy metrics for practical applications.

Ax Hadi Reisizadeh, Jiajun Ruan, Yiwei Chen, Soumyadeep Pal, Sijia Liu, Mingyi Hong 5/29/2026

Leak@$k$: Unlearning Does Not Make LLMs Forget Under Probabilistic Decoding

Research showing existing LLM unlearning methods fail under probabilistic decoding, revealing gap between deterministic and stochastic evaluation of forgotten content.

Ax Yonatan Dukler, Guihong Li, Deval Shah, Jiang Liu, Vikram Appia, Emad Barsoum 5/29/2026

FarSkip-Collective: Unhobbling Blocking Communication in Mixture of Experts Models

Architecture modifications enabling overlapping computation and communication in distributed Mixture of Experts models.

Ax Jin Huang 5/29/2026

Enhancing Reinforcement Learning in 3D Environments through Semantic Segmentation: A Case Study in ViZDoom

Semantic segmentation input representations to improve reinforcement learning efficiency in 3D environments.

Ax Ankush Kadu, Aswanth Krishnan 5/29/2026

ReflexGrad: Within-Episode Failure Recovery in LLM Agents via Progress-Gated Dual-Process Routing

Dual-process architecture enabling LLM agents to recover from failures mid-episode using gradient-based refinement.