Isolater - Feed

Ax Zhenjie Ren, Xiaoli Wei, Xiang Yu, Xun Yu Zhou 5/1/2026

Continuous-time q-learning for mean-field control with common noise, part-II: q-learning algorithms

arXiv paper on Q-learning algorithms for mean-field control with common noise in multi-agent settings.

Ax Yang Luo 5/1/2026

Proactive Dialogue Model with Intent Prediction

arXiv paper on dialogue models with intent prediction using lightweight Temporal Bayesian Networks injected via prompts.

Ax Hongliang Liu, Tung-Ling Li, Yuhao Wu 5/1/2026

Perturbation Probing: A Two-Pass-per-Prompt Diagnostic for FFN Behavioral Circuits in Aligned LLMs

Perturbation probing identifies causal circuit structures in LLM feed-forward networks across 13 models, revealing opposition and other behavioral circuits from RLHF.

Ax David Fernandez, Pedram MohajerAnsari, Amir Salarpour, Mert D. Pese 5/1/2026

Understanding Adversarial Transferability in Vision-Language Models for Autonomous Driving: A Cross-Architecture Analysis

Cross-architecture analysis of adversarial attack transferability in vision-language models for autonomous driving systems.

Ax Anna Mazhar, Huzaifa Suri, Sainyam Galhotra 5/1/2026

Trace-Level Analysis of Information Contamination in Multi-Agent Systems

Analyzes information contamination propagation in multi-agent workflows with heterogeneous data sources, showing how uncertainty redirects agent decision-making and execution paths.

Ax Ethan Bito, Yongli Ren, Estrid He 5/1/2026

One Pass, Any Order: Position-Invariant Listwise Reranking for LLM-Based Recommendation

Studies order sensitivity in LLM-based recommendation reranking and proposes position-invariant listwise ranking to fix decoder-only model vulnerabilities.

Ax Naomi Esposito, Anthony Tricarico, Luisa Porzio, Ali Aghazadeh Ardebili, Massimo Stella 5/1/2026

Math Education Digital Shadows for facilitating learning with LLMs: Math performance, anxiety and confidence in simulated students and AIs

MEDS dataset maps how 14 LLM families reason about mathematics across 28,000 personas to study model mathematical abilities and biases for education applications.

Ax Ali Aghazadeh Ardebili, Massimo Stella 5/1/2026

Mapping how LLMs debate societal issues when shadowing human personality traits, sociodemographics and social media behavior

Cognitive Digital Shadows dataset analyzing how 19 LLMs debate societal issues under different persona and sociodemographic prompts.

Ax Buqing Ou, Frederike D\"umbgen 5/1/2026

Can Tabular Foundation Models Guide Exploration in Robot Policy Learning?

TFM-S3 method uses tabular foundation models to improve global exploration in continuous control robot policy learning.

Ax Sebastiano Franchini, Alexis Carrillo, Edoardo Sebastiano De Duro, Riccardo Improta, Ali Aghazadeh Ardebili, Massimo Stella 5/1/2026

The TEA Nets framework combines AI and cognitive network science to model targets, events and actors in text

Open-source Python library extracting subjects, verbs, and objects from text using cognitive network science and AI for NLP tasks.

Ax Rog\'erio Almeida Gouv\^ea, Gian-Marco Rignanese 5/1/2026

VibroML: an automated toolkit for high-throughput vibrational analysis and dynamic instability remediation of crystalline materials using machine-learned potentials

Open-source Python toolkit using machine-learned potentials for automated phonon analysis and structural remediation of crystalline materials.

Ax Ishrak Hamim Mahi, Siam Ferdous, Md Sakib Sadman Badhon, Nabid Hasan Omi, Md Habibun Nabi Hemel, Farig Yousuf Sadeque, Md. Tanzim Reza 5/1/2026

Machine Unlearning for Class Removal through SISA-based Deep Neural Network Architectures

SISA-based architecture for machine unlearning enabling removal of class data from trained neural networks.

Ax Yuhua Wang, Qinnan Zhang, Xiaodong Li, Huan Zhang, Yifan Sun, Wangjie Qiu, Hainan Zhang, Yongxin Tong, Zhiming Zheng 5/1/2026

Taming Noise-Induced Prototype Degradation for Privacy-Preserving Personalized Federated Fine-Tuning

Privacy-preserving federated fine-tuning method addressing noise-induced prototype degradation with local differential privacy.

Ax Bowen Sun, Chaozhuo Li, Yaodong Yang, Yiwei Wang, Chaowei Xiao 5/1/2026

TwinGate: Stateful Defense against Decompositional Jailbreaks in Untraceable Traffic via Asymmetric Contrastive Learning

TwinGate defense mechanism against decompositional jailbreaks in LLMs using stateful asymmetric contrastive learning.

Ax Simon Dennis, Michael Diamond, Rivaan Patil, Kevin Shabahang, Hao Guo 5/1/2026

In-Context Prompting Obsoletes Agent Orchestration for Procedural Tasks

Empirical comparison showing in-context prompting with self-orchestration outperforms external agent frameworks for procedural tasks.

Ax Yanwu Gu, Linglong Kong, Dong Xia 5/1/2026

Prediction-powered Inference by Mixture of Experts

Mixture of Experts approach for semi-supervised inference combining multiple diverse prediction models with limited labels.

Ax Rui Xu, Yi Chen, Sihong Xie, Hui Xiong 5/1/2026

Geometry-Calibrated Conformal Abstention for Language Models

Conformal Abstention framework using conformal prediction to detect when LLMs lack knowledge and should abstain.

Ax Hanane Nour Moussa, Yifei Li, Zhuoyang Li, Yankai Yang, Cheng Tang, Tianshu Zhang, Nesreen K. Ahmed, Ali Payani, Ziru Chen, Huan Sun 5/1/2026

D3-Gym: Constructing Real-World Verifiable Environments for Data-Driven Discovery

D3-Gym dataset with 565 verifiable environments for evaluating language models and agents on scientific discovery tasks.

Ax An-Yang Ji, Jun-Peng Jiang, De-Chuan Zhan, Han-Jia Ye 5/1/2026

TopBench: A Benchmark for Implicit Prediction and Reasoning over Tabular Question Answering

TopBench benchmark for evaluating LLMs on implicit prediction and reasoning over tabular data.

Ax Sigma Jahan, Saurabh Singh Rajput, Tushar Sharma, Mohammad Masudur Rahman 5/1/2026

DEFault++: Automated Fault Detection, Categorization, and Diagnosis for Transformer Architectures

DEFault++ automated fault detection system for identifying component-level failures in transformer architectures.

Ax Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, Rossano Venturini 5/1/2026

Efficient Multivector Retrieval with Token-Aware Clustering and Hierarchical Indexing

Token-aware clustering and hierarchical indexing method for efficient multivector retrieval models used in NLP applications.

Ax Daniel Waxman, Fernando Llorente, Petar M. Djuri\'c 5/1/2026

Sequential Inference for Gaussian Processes: A Signal Processing Perspective

Sequential inference methods for Gaussian Processes from signal processing perspective, adapting ML models for streaming/online scenarios.

Ax Grace T. Bai, Brandon B. Le 5/1/2026

Mapping the Phase Diagram of the Vicsek Model with Machine Learning

ML classification of Vicsek flocking model phase diagram using K-means clustering on dynamical observables across parameter space.

Ax Sriram Narayanan, Ziyu Jiang, Srinivasa Narasimhan, Manmohan Chandraker 5/1/2026

PhyCo: Learning Controllable Physical Priors for Generative Motion

PhyCo framework adds physical constraints to video diffusion models for realistic object motion, collision physics, and material responses in generated videos.

Ax Tao Ge, Baolin Peng, Hao Cheng, Jianfeng Gao 5/1/2026

Synthetic Computers at Scale for Long-Horizon Productivity Simulation

Scalable methodology for generating synthetic computer environments with realistic folder hierarchies and artifacts for productivity task simulation.

Ax Chengyang Huang, Siddhartha Srivastava, Kenneth K. Y. Ho, Kathy E. Luker, Gary D. Luker, Xun Huan, Krishna Garikipati 5/1/2026

FP-IRL: Fokker--Planck Inverse Reinforcement Learning -- A Physics-Constrained Approach to Markov Decision Processes

Physics-constrained inverse reinforcement learning using Fokker-Planck equation to infer reward functions from trajectories without known dynamics.

Ax Lu Yin, Ajay Jaiswal, Shiwei Liu, Souvik Kundu, Zhangyang Wang 5/1/2026

Junk DNA Hypothesis: Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs "Difficult" Downstream Tasks in LLMs

Analysis of LLM pruning showing small-magnitude pre-trained weights are critical for downstream task performance, contradicting redundancy assumptions.

Ax David J. Schodt, Ryan Brown, Michael Merritt, Samuel Park, Delsin Menolascino, Mark A. Peot 5/1/2026

A Framework for Variational Inference of Lightweight Bayesian Neural Networks with Heteroscedastic Uncertainties

Framework for variational inference in Bayesian neural networks with heteroscedastic uncertainty estimation.

Ax Pavlin G. Poli\v{c}ar, Bla\v{z} Zupan 5/1/2026

VERA: Generating Visual Explanations of Two-Dimensional Embeddings via Region Annotation

VERA system for generating visual explanations of 2D embeddings from dimensionality reduction techniques via region annotation.

Ax Mingyuan Wang, Yangzi Guo, Sida Liu, Yuhang Liu 5/1/2026

Exploring Vision Neural Network Pruning via Screening Methodology

Network pruning framework for reducing parameters and computational costs in deep neural networks for edge device deployment.

Ax Alessandro Daniele, Emile van Krieken 5/1/2026

Gradient-Based Optimization on G\"odel Logic as Discrete Local Search

Gradient-based optimization on Gödel logic for neurosymbolic systems, enabling discrete logical reasoning with continuous optimization.

Ax Francesco D'Amico, Dario Bocchi, Matteo Negri 5/1/2026

Implicit bias produces neural scaling laws in learning curves, from perceptrons to deep networks

Theoretical analysis of scaling laws in deep learning through implicit bias, explaining power-law relationships between model performance and resource growth.

Ax Perry Dong, Qiyang Li, Dorsa Sadigh, Chelsea Finn 5/1/2026

EXPO: Stable Reinforcement Learning with Expressive Policies

Training expressive RL policies like diffusion and flow-matching models using online RL with offline data, addressing gradient stability challenges in complex policy parameterizations.

Ax Zachary Robertson, Sanmi Koyejo 5/1/2026

Let's Measure Information Step-by-Step: AI-Based Evaluation Beyond Vibes

arXiv paper on evaluating AI systems without ground truth using information theory and mutual information, addressing adversarial manipulation in agent evaluation.

Ax Zhengyan Huan, Jacob Boerma, Li-Ping Liu, Shuchin Aeron 5/1/2026

Constraint-Aware Flow Matching via Randomized Exploration

arXiv paper on constraint-aware flow matching for generative models addressing constraint violations.

Ax Hugh Xuechen Liu, K{\i}van\c{c} Tatar 5/1/2026

AEGIS: Authentic Edge Growth In Sparsity for Link Prediction in Edge-Sparse Bipartite Knowledge Graphs

arXiv paper presenting AEGIS, an edge augmentation framework for link prediction in sparse bipartite knowledge graphs.

Ax Lute Lillo, Nick Cheney 5/1/2026

Activation Function Design Sustains Plasticity in Continual Learning

arXiv paper on activation function design's role in preventing plasticity loss during continual learning.

Ax Aidan Acquah, Shing Chan, Aiden Doherty 5/1/2026

ActiNet: An Open-Source Tool for Activity Intensity Classification of Wrist-Worn Accelerometry Using Self-Supervised Deep Learning

ActiNet open-source tool using self-supervised deep learning and HMMs for activity intensity classification from wrist accelerometry.

Ax Lorenzo Nikiforos, Luciano Prono, Charalampos Antoniadis, Fabio Pareschi, Riccardo Rovatti, Gianluca Setti 5/1/2026