Isolater - Feed

Ax Zhenyu Liao, Michael W. Mahoney 4/17/2026

Random Matrix Theory for Deep Learning: Beyond Eigenvalues of Linear Models

Theoretical analysis extending random matrix theory to understand behavior of overparameterized deep neural networks in high-dimensional settings.

Ax Baochang Ren, Shuofei Qiao, Da Zheng, Huajun Chen, Ningyu Zhang 4/17/2026

KnowRL: Exploring Knowledgeable Reinforcement Learning for Factuality

KnowRL reinforcement learning approach to reduce hallucination in LLMs by providing factual supervision during reasoning process.

Ax Parv Kapoor, Akila Ganlath, Michael Clifford, Changliu Liu, Sebastian Scherer, Eunsuk Kang 4/17/2026

Constrained Decoding for Safe Robot Navigation Foundation Models

Constrained decoding approach for ensuring safe navigation in robotic foundation models with explicit behavioral constraints.

Ax Haonan Chen, Jiaming Xu, Hongyu Chen, Kaiwen Hong, Binghao Huang, Chaoqi Liu, Jiayuan Mao, Yunzhu Li, Yilun Du, Katherine Driggs-Campbell 4/17/2026

Multi-Modal Manipulation via Multi-Modal Policy Consensus

Multi-modal policy consensus approach factorizing robotic manipulation policies into modality-specific diffusion models for flexible integration.

Ax Tianyu Li, Yihan Li, Zizhe Zhang, Nadia Figueroa 4/17/2026

Flow with the Force Field: Learning 3D Compliant Flow Matching Policies from Force and Demonstration-Guided Simulation Data

Flow matching policies for contact-rich robotic manipulation tasks using force and demonstration-guided simulation data.

Ax Austin Spizzirri 4/17/2026

The Specification Trap: Why Static Value Alignment Alone Is Insufficient for Robust Alignment

Philosophical analysis arguing static value alignment insufficient for robust AI alignment under capability scaling and distributional shift.

Ax Garud Iyengar, Yu-Shiou Willy Lin, Kaizheng Wang 4/17/2026

Model-Free Assessment of Simulator Fidelity via Quantile Curves

Model-free methods for assessing simulator fidelity and quantifying sim-to-real gaps in generative AI models.

Ax Rei Taniguchi, Yuyang Dong, Makoto Onizuka, Chuan Xiao 4/17/2026

Adaptive Layer Selection for Layer-Wise Token Pruning in LLM Inference

Adaptive layer selection technique for layer-wise token pruning to reduce KV cache in LLM inference with improved efficiency.

Ax Lirui Zhang, Huishuai Zhang 4/17/2026

De-Anonymization at Scale via Tournament-Style Attribution

DAS method for authorship de-anonymization at scale using tournament-style attribution with LLMs, analyzing privacy implications.

Ax Sukumar Kishanthan, Kumar Thushalika, Buddhi Jayasekara, Asela Hevapathige 4/17/2026

Beyond Translation: Evaluating Mathematical Reasoning Capabilities of LLMs in Sinhala and Tamil

Evaluation of LLM mathematical reasoning capabilities in low-resource languages Sinhala and Tamil.

Ax Drew Prinster, Clara Fannjiang, Ji Won Park, Kyunghyun Cho, Anqi Liu, Suchi Saria, Samuel Stanton 4/17/2026

Conformal Policy Control

Conformal policy control framework using safe reference policies as regulators for untested policies in high-stakes agent exploration.

Ax Markus Knauer, Samuel Bustamante, Thomas Eiband, Alin Albu-Sch\"affer, Freek Stulp, Jo\~ao Silv\'erio 4/17/2026

IROSA: Interactive Robot Skill Adaptation using Natural Language

IROSA framework combining foundation models with imitation learning for robot skill adaptation via natural language instructions.

Ax Shiwan Zhao, Zhihu Wang, Xuyang Zhao, Jiaming Zhou, Caiyue Xu, Chenfei Liu, Liting Zhang, Yuhang Jia, Yanzhe Zhang, Hualong Yu, Zichen Xu, Qicheng Li, Yong Qin 4/17/2026

Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning

Survey unifying LLM post-training methods (SFT, preference optimization, RL, process supervision) through lens of off-policy and on-policy learning.

Ax Utshab Kumar Ghosh, Ashish David, Shubham Chatterjee 4/17/2026

Reproduction Beyond Benchmarks: ConstBERT and ColBERT-v2 Across Backends and Query Distributions

Reproducibility study of ColBERT-v2 and ConstBERT showing 86-97% performance drop on long narrative queries due to architectural limitations.

Ax Imanol Miranda, Ander Salaberria, Eneko Agirre, Gorka Azkune 4/17/2026

Revisiting Compositionality in Dual-Encoder Vision-Language Models: The Role of Inference

Research on CLIP and dual-encoder vision-language models showing compositional limitations stem from inference protocol, not representations.

Ax Hiranya V. Peiris 4/17/2026

Functional Emotions or Situational Contexts? A Discriminating Test from the Mythos Preview System Card

Research using SAE features and emotion vectors to study Claude Mythos Preview system behavior during misaligned responses.

Ax Itay Itzhak, Eliya Habba, Gabriel Stanovsky, Yonatan Belinkov 4/17/2026

From Feelings to Metrics: Understanding and Formalizing How Users Vibe-Test LLMs

Research formalizing user 'vibe-testing' practices for evaluating LLMs, moving from informal experience-based evaluation to structured methodology for reproducibility.

Ax Xingyu Xiao, Jiejuan Tong, Jun Sun, Zhe Sui, Peng Chen, Jingang Liang, Haitao Wang 4/17/2026

NuHF Claw: A Risk Constrained Cognitive Agent Framework for Human Centered Procedure Support in Digital Nuclear Control Rooms

Research paper proposing NuHF Claw cognitive agent framework using LLMs for safety-critical decision support in nuclear control rooms.

Ax Hong Su 4/17/2026

Simulating Human Cognition: Heartbeat-Driven Autonomous Thinking Activity Scheduling for LLM-based AI systems

Heartbeat-driven autonomous thinking framework for LLM agents enabling proactive scheduling and continuous reflection instead of reactive control flows.

Ax Pierre Lotte (EPE UT, IRIT), Andr\'e P\'eninou (UT2J, IRIT-SIG, IRIT), Olivier Teste (IRIT-SIG, IRIT, UT2J, Comue de Toulouse) 4/17/2026

Fun-TSG: A Function-Driven Multivariate Time Series Generator with Variable-Level Anomaly Labeling

Synthetic multivariate time series generator with fine-grained anomaly annotations and variable-level dependencies for benchmarking detection methods.

Ax Pramudita Satria Palar, Paul Saves, Muhammad Daffa Robani, Nicolas Verstaevel, Moncef Garouani, Julien Aligon, Koji Shimoyama, Joseph Morlier, Benoit Gaudou 4/17/2026

Interpretable and Explainable Surrogate Modeling for Simulations: A State-of-the-Art Survey and Perspectives on Explainable AI for Decision-Making

Survey of interpretable surrogate modeling for complex system simulations, covering explainable AI techniques for decision-making and black-box model transparency.

Ax Wangjie Gan, Miao Pan, Linbo Xi, Wenqi Zhang, Jintao Chen, Jianwei Yin, Xuhong Zhang 4/17/2026

GFT: From Imitation to Reward Fine-Tuning with Unbiased Group Advantages and Dynamic Coefficient Rectification

Training dynamics analysis unifying supervised fine-tuning and reinforcement learning for LLMs via reward perspective, with group advantage optimization and coefficient rectification.

Ax Kinhei Lee, Peiyuan Jing, Zhenxuan Zhang, Yue Yang, Tao Wang, Dominic C Marshall, Yingying Fang, Guang Yang 4/17/2026

Seeing Through Experts Eyes A Foundational Vision Language Model Trained on Radiologists Gaze and Reasoning

Vision-language model trained on radiologist gaze and reasoning patterns to improve chest X-ray interpretation by emulating expert diagnostic workflows.

Ax Aaron Pache, Mark CW van Rossum 4/17/2026

Mistake gating leads to energy and memory efficient continual learning

Biologically-inspired mistake-gated learning mechanism reducing synaptic plasticity costs, achieving energy and memory efficient continual learning in neural networks.

Ax Duo Lu, Andrew Crotty, U\u{g}ur \c{C}etintemel 4/17/2026

Credo: Declarative Control of LLM Pipelines via Beliefs and Policies

Framework for declarative control of LLM agent pipelines using beliefs and policies, enabling transparent behavior adaptation and stateful decision-making.

Ax Ivan Ternovtsii, Yurii Bilak 4/17/2026

Equifinality in Mixture of Experts: Routing Topology Does Not Determine Language Modeling Quality

Geometric MoE architecture using cosine-similarity routing in low-dimensional space, achieving 80% fewer routing parameters while maintaining language modeling quality.

Ax Muhammad Imam Luthfi Balaka, Raul Castro Fernandez 4/17/2026

Demonstration of Pneuma-Seeker: Agentic System for Reifying and Fulfilling Information Needs on Tabular Data

Agentic system for interactive data exploration that reifies vague information needs into explicit relational specifications with iterative refinement and provenance tracking.

Ax Ivan Ternovtsii, Yurii Bilak 4/17/2026

Geometric Routing Enables Causal Expert Control in Mixture of Experts

Demonstrates causal meaningfulness of individual experts in sparse MoE models via geometric routing, showing monosemantic expert behavior despite routing topology equivalence.

Ax Ruiyi Zhang, Peijia Qin, Qi Cao, Li Zhang, Pengtao Xie 4/17/2026

AIBuildAI: An AI Agent for Automatically Building AI Models

AutoML system that automates AI model development including architecture design, feature engineering, training pipeline implementation, and empirical refinement.

Ax Saumik Narayanan, Raja Panjwani, Siddhartha Sen, Chien-Ju Ho 4/17/2026

Improving Human Performance with Value-Aware Interventions: A Case Study in Chess

Framework for value-aware AI interventions in sequential decision-making, accounting for human execution limitations rather than assuming optimal follow-up actions.

Ax Jillian Fisher, Jennifer Neville, Chan Young Park 4/17/2026

Response-Aware User Memory Selection for LLM Personalization

Novel method for personalizing LLMs by selecting user memory based on response utility rather than similarity, optimizing model behavior at inference.

Ax Weixiang Shen, Bailiang Jian, Jun Li, Che Liu, Johannes Moll, Xiaobin Hu, Daniel Rueckert, Hongwei Bran Li, Jiazhen Pan 4/17/2026

Evo-MedAgent: Beyond One-Shot Diagnosis with Agents That Remember, Reflect, and Improve

LLM agent system for medical diagnosis that accumulates experience across cases, reflects on mistakes, and improves tool-use behavior without reinforcement learning.

Ax Nina \.Zukowska, Wolfgang Stammer, Bernt Schiele, Jonas Fischer 4/17/2026

Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformers

Mechanistic interpretability approach for Vision Transformers using circuit-based analysis similar to LLM interpretability, improving model transparency.

Ax Nenad Banfic, David Fan, Kunal Vaishnavi, Sam Kemp, Sunghoon Choi, Rui Ren, Sayan Shaw, Meng Tang 4/17/2026

Pushing the Limits of On-Device Streaming ASR: A Compact, High-Accuracy English Model for Low-Latency Inference

Empirical study optimizing automatic speech recognition models for on-device CPU inference across encoder-decoder, transducer, and LLM-based architectures.

Ax Mel Sohm, Charles Dezons, Sami Sellami, Oscar Ninou, Axel Pincon 4/17/2026

Improving Machine Learning Performance with Synthetic Augmentation

Formalizes synthetic data augmentation as distribution modification in financial ML, analyzing bias-variance tradeoffs from statistical perspective.

Ax Dongxin Guo, Jikun Wu, Siu Ming Yiu 4/17/2026

Geometric Metrics for MoE Specialization: From Fisher Information to Early Failure Detection

Information-geometric framework for characterizing Mixture-of-Experts specialization dynamics using Fisher information, addressing limitations of existing metrics.

Ax Michal Rosen-Zvi, Yoav Kan-Tor, Michael Danziger, Agata Ferretti, Javier Aula-Blasco, Julia Falcao, Ron Shamir, Mordechai Muszkat 4/17/2026

Perspective on Bias in Biomedical AI: Preventing Downstream Healthcare Disparities

Perspective on sources of bias in biomedical AI during data collection and research prioritization stages.

Ax MindDR Team, Li Auto Inc 4/17/2026

Mind DeepResearch Technical Report

Mind DeepResearch efficient multi-agent framework with planning, search, and report agents achieving strong performance with 30B-parameter models.

Ax Rohit Kumar Salla, Ramya Manasa Amancherla, Manoj Saravanan 4/17/2026

Quantifying Cross-Query Contradictions in Multi-Query LLM Reasoning

Benchmark and metrics for measuring logical consistency in multi-query LLM reasoning with entailment and contradiction analysis.

Ax Wei Zhu, Jian Zhang, Lixing Yu, Kun Yue, Zhiwen Tang 4/17/2026

Dissecting Failure Dynamics in Large Language Model Reasoning

Analysis of LLM reasoning failures showing errors originate from early transition points before coherent but incorrect local reasoning.

Ax Adam Rida 4/17/2026

TRACER: Trace-Based Adaptive Cost-Efficient Routing for LLM Classification

TRACER uses lightweight surrogates trained on production logs to adaptively route LLM classification traffic for cost efficiency.

Ax Pengfei Li, Shijie Wang, Fangyuan Li, Yikun Fu, Kaifeng Liu, Kaiyan Zhang, Dazhi Zhang, Yuqiang Li, Biqing Qi, Bowen Zhou 4/17/2026

MARS$^2$: Scaling Multi-Agent Tree Search via Reinforcement Learning for Code Generation

MARS² combines reinforcement learning with multi-agent tree search to improve trajectory diversity and performance in code generation.

Ax Md Arid Hasan, Azhagu Meena SP, Aditya Khan, Abu Md Akteruzzaman Bhuiyan, Helal Uddin Ahmed, Joysree Debi, Farig Sadeque, Annie En-Shiun Lee, Syed Ishtiaque Ahmed 4/17/2026

Enhancing Mental Health Counseling Support in Bangladesh using Culturally-Grounded Knowledge

RAG and fine-tuning approaches to improve LLM cultural sensitivity and clinical appropriateness for Bangladesh mental health counseling.

Ax Xing Zhang, Guanghui Wang, Yanwei Cui, Wei Qiu, Ziyuan Li, Bing Zhu, Peiyang He 4/17/2026

Prompt Optimization Is a Coin Flip: Diagnosing When It Helps in Compound AI Systems

Statistical analysis of prompt optimization effectiveness across compound AI systems, identifying when methods succeed or fail.

Ax Ha Thanh Nguyen, Wachara Fungwacharakorn, Sabine Wehnert, May Myo Zin, Yuntao Kong, Jieying Xue, Micha{\l} Araszkiewicz, Randy Goebel, Ken Satoh 4/17/2026

GDPR Auto-Formalization with AI Agents and Human Verification

Human-in-the-loop multi-agent system for automatic GDPR formalization using LLMs with iterative feedback and verification modules.

Ax Zijian Zhang, Aiwei Yin, Amaan Baweja, Jiaru Bai, Ignacio Gustin, Varinia Bernales, Al\'an Aspuru-Guzik 4/17/2026

El Agente Forjador: Task-Driven Agent Generation for Quantum Simulation

El Agente Forjador multi-agent framework for quantum simulation where universal coding agents autonomously generate domain-specific tools.

Ax Yubin Kim, Salman Rahman, Samuel Schmidgall, Chunjong Park, A. Ali Heydari, Ahmed A. Metwally, Hong Yu, Xin Liu, Xuhai Xu, Yuzhe Yang, Maxwell A. Xu, Zhihan Zhang, Cynthia Breazeal, Tim Althoff, Petar Sirkovic, Ivor Rendulic, Annalisa Pawlosky, Nicolas Stroppa, Juraj Gottweis, Elahe Vedadi, Alan Karthikesalingam, Pushmeet Kohli, Vivek Natarajan, Mark Malhotra, Shwetak Patel, Hae Won Park, Hamid Palangi, Daniel McDuff 4/17/2026