Isolater - Feed

Ax Wenbing Li, Zikai Song, Hang Zhou, Yunyao Zhang, Junqing Yu, Wei Yang 5/14/2026

LoRA-Mixer: Coordinate Modular LoRA Experts Through Serial Attention Routing

LoRA-Mixer: modular mixture-of-experts framework routing task-specific LoRA experts through attention matrices for efficient multi-task LLM adaptation.

Ax Shivansh Patel, Shraddhaa Mohan, Hanlin Mai, Unnat Jain, Svetlana Lazebnik, Yunzhu Li 5/14/2026

Robotic Manipulation by Imitating Generated Videos Without Physical Demonstrations

RIGVid: system enabling robots to learn complex manipulation tasks by imitating AI-generated videos without physical demonstrations.

Ax John Wu, Adam Cross, Jimeng Sun 5/14/2026

RDMA: Cost Effective Agent-Driven Rare Disease Mining from Electronic Health Records

Agent-driven system for mining rare disease information from unstructured clinical notes using LLMs to improve medical documentation.

Ax Yifan Zhang 5/14/2026

A Markov Categorical Framework for Language Modeling

Analytical framework modeling autoregressive language models as composition of information-processing stages using Markov categories.

Ax Minghao Liu, Chia-Hsuan Lu, Marta Kwiatkowska 5/14/2026

Exact Verification of Graph Neural Networks with Incremental Constraint Solving

Exact verification method for graph neural networks using incremental constraint solving to provide adversarial robustness guarantees.

Ax Amirmohsen Sattarifard, Sepehr Lavasani, Kunlin Zhang, Amirhossein Rajabpour, Hanlin Xu, Fengyu Sun, Negar Hassanpour, Chao Gao 5/14/2026

GLASS: Global-Local Aggregation for Inference-time Sparsification of LLMs

GLASS: inference-time sparsification method for LLMs using global-local aggregation to improve deployment on resource-constrained devices.

Ax Dikshant Sagar, Kaiwen Yu, Alejandro Yankelevich, Jianming Bian, Pierre Baldi 5/14/2026

Adapting Vision-Language Models for Neutrino Event Classification in High-Energy Physics

Fine-tuned LLaMA 3.2 vision-language model adapted for neutrino event classification in high-energy physics detector data.

Ax Zongyue Li, Xiao Han, Yusong Li, Niklas Strauss, Matthias Schubert 5/14/2026

DAWM: Diffusion Action World Models for Offline Reinforcement Learning via Action-Inferred Transitions

Diffusion-based world models for offline reinforcement learning that generate actions alongside states and rewards for improved TD learning compatibility.

Ax Zheng Wang, Kaixuan Zhang, Wanfang Chen, Xiaonan Lu, Longyuan Li, Tobias Schlagenhauf 5/14/2026

Characteristic Root Analysis and Regularization for Linear Time Series Forecasting

Theoretical study of linear models for time series forecasting, analyzing their robustness and interpretability through characteristic root analysis.

Ax Danial Samadi Vahdati, Tai Duc Nguyen, Ekta Prashnani, Koki Nagano, David Luebke, Orazio Gallo, Matthew Stamm 5/14/2026

Unmasking Puppeteers: Leveraging Biometric Leakage to Expose Impersonation in AI-Based Videoconferencing

Security method detecting impersonation in AI videoconferencing by exploiting biometric leakage from pose-expression latents to expose puppeteering attacks.

Ax Jehyeok Yeon, Isha Chaudhary, Gagandeep Singh 5/14/2026

Quantitative Certification of Agentic Tool Selection

Framework for certifying tool selection in LLM-based agentic systems, evaluating robustness against adversarial tool pools and deployment scenarios.

Ax Jianke Zhang, Yucheng Hu, Yanjiang Guo, Xiaoyu Chen, Yichen Liu, Wenna Chen, Chaochao Lu, Jianyu Chen 5/14/2026

UniJEPA: Enhancing Robot Policy via Unified Continuous and Discrete Representation Learning

Robot policy learning method combining continuous and discrete representations using vision-language and visual dynamics models for diverse manipulation tasks.

Ax Mike Lasby, Ivan Lazarevich, Nish Sinnadurai, Sean Lie, Yani Ioannou, Vithursan Thangarasa 5/14/2026

REAP the Experts: Why Pruning Prevails for One-Shot MoE compression

Study comparing expert pruning vs merging strategies for compressing Mixture-of-Experts models, finding pruning superior for generative tasks.

Ax Luca Belli, Kate H. Bentley, Will Alexander, Emily Ward, Matt Hawrilenko, Kelly Johnston, Mill Brown, Adam M. Chekroud 5/14/2026

VERA-MH Concept Paper

Automated safety evaluation framework for mental health AI chatbots using clinician-informed rubric and multi-agent validation approach.

Ax Dario Shariatian, Alain Durmus, Umut Simsekli, Stefano Peluchetti 5/14/2026

Latent-Augmented Discrete Diffusion Models

Latent-augmented discrete diffusion model with auxiliary channel for improved few-step language generation and cross-token dependencies.

Ax Song Bian, Tao Yu, Shivaram Venkataraman, Youngsuk Park 5/14/2026

Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs

Study on trade-offs between model accuracy and inference efficiency examining architectural factors affecting LLM deployment costs.

Ax Yi He, Xingyu Zhou 5/14/2026

On the Sample Complexity of Differentially Private Policy Optimization

Theoretical analysis of sample complexity in differentially private policy optimization for reinforcement learning with privacy guarantees.

Ax John Yang, Kilian Lieret, Joyce Yang, Carlos E. Jimenez, Muhtasham Oblokulov, Aryan Siddiqui, Ofir Press, Ludwig Schmidt, Diyi Yang 5/14/2026

CodeClash: Benchmarking Goal-Oriented Software Engineering

Benchmark evaluating whether LLMs can iteratively develop code toward high-level goals beyond isolated task completion.

Ax Zirui Zhao, Boye Niu, Harold Soh, David Hsu, Wee Sun Lee 5/14/2026

When Diffusion Breaks Constraints: Sequential Autoregressive Generation with RL and MCTS

Sequential autoregressive generation with reinforcement learning and MCTS to enforce hard constraints in planning and design tasks.

Ax Daeun Lee, Subhojyoti Mukherjee, Branislav Kveton, Ryan A. Rossi, Viet Dac Lai, Seunghyun Yoon, Trung Bui, Franck Dernoncourt, Mohit Bansal 5/14/2026

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

Benchmark for evaluating multimodal LLMs on streaming video understanding with human gaze signals for AR applications.

Ax Ziqing Wen, Jiahuan Wang, Ping Luo, Dongsheng Li, Tao Sun 5/14/2026

FOAM: Blocked State Folding for Memory-Efficient LLM Training

Memory-efficient LLM training technique using blocked state folding to reduce memory bottlenecks with Adam optimizer.

Ax Zihan Yao, Ruoyu Wu, Tianxiang Gao 5/14/2026

Feature Learning Dynamics in Infinite-Depth Neural Networks

Theoretical analysis of feature evolution dynamics in infinite-depth ResNets under depth-μP scaling during training.

Ax Yifan Zhang, Yifeng Liu, Mengdi Wang, Quanquan Gu 5/14/2026

Deep Delta Learning

Deep Delta Learning introduces residual update rule for transformer layers enabling selective content rewriting in deep networks.

Ax Yinzhu Chen, Abdine Maiga, Hossein A. Rahmani, Emine Yilmaz 5/14/2026

Automated Rubrics for Reliable Evaluation of Medical Dialogue Systems

Multi-agent framework using retrieval-augmented generation and LLM judges to create automated rubrics for evaluating medical dialogue systems.

Ax Or Ordentlich, Yury Polyanskiy 5/14/2026

High-Rate Quantized Matrix Multiplication I

Research on quantized matrix multiplication optimization for efficient LLM deployment with weight and activation quantization.

Ax Yanxi Wang, Zhiling Zhang, Wenbo Zhou, Weiming Zhang, Jie Zhang, Qiannan Zhu, Yu Shi, Shuxin Zheng, Jiyan He 5/14/2026

GUIGuard-Bench: Toward a General Evaluation for Privacy-Preserving GUI Agents

Benchmark evaluating privacy risks in GUI agents that process screenshots, measuring exposure of sensitive information during operation.

Ax Adam Remaki, Christel G\'erardin, Eul\`alia Farr\'e-Maduell, Martin Krallinger, Xavier Tannier 5/14/2026

SynCABEL: Synthetic Contextualized Augmentation for Biomedical Entity Linking

Framework using LLMs to generate synthetic training data for biomedical entity linking, addressing scarcity of expert annotations.

Ax Jian-Feng Cai, Haixia Liu, Zhengyi Su, Chao Wang 5/14/2026

Improving Classifier-Free Guidance of Flow Matching via Manifold Projection

Principled optimization-based interpretation of classifier-free guidance in flow matching models for controllable generation tasks.

Ax Puyu Wang, Junyu Zhou, Philipp Liznerski, Marius Kloft 5/14/2026

Optimization, Generalization and Differential Privacy Bounds for Gradient Descent on Kolmogorov-Arnold Networks

Theoretical analysis of gradient descent training dynamics, generalization bounds, and differential privacy properties for Kolmogorov-Arnold Networks (KANs).

Ax Hasi Hays, William J. Richardson 5/14/2026

RAG-GNN: Integrating Retrieved Knowledge with Graph Neural Networks for Precision Medicine

RAG-GNN framework combines graph neural networks with retrieved biomedical literature for cancer signaling predictions using contrastive alignment and gated fusion mechanisms.

Ax Yoonjun Cho, Dongjae Jeon, Soeun Kim, Moongyu Jeon, Albert No 5/14/2026

Preserve-Then-Quantize: Balancing Rank Budgets for Quantization Error Reconstruction in LLMs

Post-training quantization method balancing rank budgets between preserving low-rank structure and reconstructing quantization error in LLMs.

Ax Ibrahim Albool, Malak Gamal El-Din, Salma Elmalaki, Yasser Shoukry 5/14/2026

Exponential Approximation Rates and Parameter Efficiency of Learnable Bernstein Activations

Theoretical analysis of learnable Bernstein polynomial activations showing exponential approximation rates and parameter efficiency.

Ax Atula Tejaswi, Litu Rout, Constantine Caramanis, Sanjay Shakkottai, Sujay Sanghavi 5/14/2026

Entropy Aware Reward Guidance for Diffusion Language Model Alignment

Entropy-aware reward guidance mechanism for test-time adaptation in discrete diffusion language models.

Ax Vishak K Bhat, Prateek Chanda, Vijval Ekbote, Ashmit Khandelwal, Maitreyi Swaroop, Vineeth N. Balasubramanian, Subbarao Kambhampati, Nagarajan Natarajan, Amit Sharma 5/14/2026

interwhen: A Generalizable Framework for Steering Reasoning Models with Test-time Verification

Single-trajectory verification framework that steers reasoning models by providing intermediate feedback during inference.

Ax Junchen Liu, Sven Elflein, Or Litany, Zan Gojcic, Ruilong Li 5/14/2026

Test-Time Training with KV Binding Is Secretly Linear Attention

Analysis revealing test-time training with KV binding is equivalent to learned linear attention, challenging memorization interpretation.

Ax Gaia Molinaro, Dave August, Danielle Perszyk, Anne G. E. Collins 5/14/2026

Language Model Goal Selection Differs from Humans' in a Self-Directed Learning Task

Empirical study showing LLMs differ from humans in goal selection during self-directed learning, challenging their use as proxies for human preferences.

Ax Leo Benac, Jonas Raedler, Zilin Ma, Finale Doshi-Velez 5/14/2026

A Benchmark for Multi-Party Negotiation Games from Real Negotiation Data

Benchmark and framework for multi-party negotiation games with binding action-level commitments derived from real climate negotiation data.

Ax Tomas Ruiz, Zhen Qin, Yifan Zhang, Xuyang Shen, Yiran Zhong, Mengdi Wang 5/14/2026

FlashSampling: Fast and Memory-Efficient Exact Sampling

Optimized sampling primitive that fuses categorical sampling into LM-head computation, reducing memory traffic for large-vocabulary decoding.

Ax Yu-Ning Qiu, Lin-Feng Zou, Jiong-Da Wang, Xue-Rong Yuan, Wang-Zhou Dai 5/14/2026

Procedural Refinement by LLM-driven Algorithmic Debugging for ARC-AGI-2

LLM-driven algorithmic debugging approach for abstract reasoning tasks using abduction-based program refinement.

Ax Ian Osband 5/14/2026

Delightful Distributed Policy Gradient

Distributed reinforcement learning method addressing negative learning from high-surprisal data in stale actor settings.

Ax Kai-Wei Chang, Wei-Chih Chen, En-Pei Hu, Hung-yi Lee, James Glass 5/14/2026

TiCo: Time-Controllable Spoken Dialogue Model

Spoken dialogue model with controllable response duration for voice assistants and interactive agents.

Ax Xiang Li, Yixuan Jia, Xiao Li, Jeffrey A. Fessler, Rongrong Wang, Qing Qu 5/14/2026

MCLR: Improving Conditional Modeling via Inter-Class Likelihood-Ratio Maximization and Unifying Classifier-Free Guidance with Alignment Objectives

Theoretical analysis of classifier-free guidance in diffusion models, studying why inference-time guidance is necessary for generative modeling.

Ax Fangxin Wang, Peyman Baghershahi, Langzhou He, Henry Peng Zou, Sourav Medya, Philip S. Yu 5/14/2026