Isolater - Feed

Ax Armin Karamzade, Kyungmin Kim, JB Lanier, Davide Corsi, Roy Fox 4/17/2026

Model-Based Reinforcement Learning under Random Observation Delays

Studies reinforcement learning under random sensor delays in POMDPs where observations arrive out-of-sequence, addressing real-world RL challenges.

Ax Vladimir Fanaskov, Vladislav Trifonov, Alexander Rudikov, Ekaterina Muravleva, Ivan Oseledets 4/17/2026

Deep Learning for Subspace Regression

Research on reduced-order modeling using deep learning to compute linear subspaces for parametric systems with offline/online stages.

Ax Enda Yu, Dezun Dong, Zhaoning Zhang, Zhe Bai, Weiling Yang, Haojie Wang, Dongsheng Li, Yongwei Wu, Xiangke Liao 4/17/2026

LayerScope: Predictive Cross-Layer Scheduling for Efficient Multi-Batch MoE Inference on Legacy Servers

Presents PreScope, a prediction-driven scheduling system for efficient MoE inference on commodity hardware with CPU offloading.

Ax Gunjun Lee, Jiwon Kim, Jaiyoung Park, Younjoo Lee, Jung Ho Ahn 4/17/2026

From Tokens to Layers: Redefining Stall-Free Scheduling for MoE Serving with Layered Prefill

Proposes layered prefill scheduling for MoE LLM inference to optimize time-to-first-token and throughput while managing compute/memory constraints.

Ax Wenshuo Wang, Ziyou Jiang, Junjie Wang, Mingyang Li, Jie Huang, Yuekai Huang, Zhiyuan Chang, Feiyan Duan, Qing Wang 4/17/2026

Fall into a Pit, Gain in a Wit: Cognitive-Guided Harmful Meme Detection via Misjudgment Risk Pattern Retrieval

PatMD approach for detecting harmful memes by learning from misjudgment patterns in multimodal content with implicit rhetorical devices.

Ax Harris Hardiman-Mostow, Jack Mauro, Adrien Weihs, Andrea L. Bertozzi 4/17/2026

Topology-Aware Active Learning on Graphs

Graph-topological active learning using Balanced Forman Curvature for coreset construction under label budget constraints.

Ax Chenming Tang, Hsiu-Yuan Huang, Weijie Liu, Clive Bai, Saiyong Yang, Yunfang Wu 4/17/2026

Do Not Step Into the Same River Twice: Learning to Reason from Trial and Error

Reinforcement learning approach for language model reasoning that learns from trial-and-error to overcome exploration stagnation in RLVR.

Ax Zhuoyun Du, Runze Wang, Huiyu Bai, Zouying Cao, Xiaoyong Zhu, Yu Cheng, Bo Zheng, Wei Chen, Haochao Ying 4/17/2026

Enabling Agents to Communicate Entirely in Latent Space

Interlat enables LLM-based agents to communicate in latent space instead of natural language, improving information transfer depth.

Ax Genghan Zhang, Shaowei Zhu, Anjiang Wei, Zhenyu Song, Allen Nie, Zhen Jia, Nandita Vijaykumar, Yida Wang, Kunle Olukotun 4/17/2026

AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimization

AccelOpt is a self-improving LLM agent that autonomously optimizes kernels for AI accelerators using iterative generation and optimization memory.

Ax Yegon Kim, Hyunsu Kim, Gyeonghoon Ko, Juho Lee 4/17/2026

Active Learning with Selective Time-Step Acquisition for PDEs

Active learning framework for PDE surrogate modeling with selective time-step acquisition to reduce training data generation costs.

Ax Qiwei Tian, Chenhao Lin, Zhengyu Zhao, Chao Shen 4/17/2026

Pay Less Attention to Function Words for Free Robustness of Vision-Language Models

Function-word De-Attention method improves robustness of vision-language models against cross-modal adversarial attacks.

Ax Jeff J. Ma, Jae-Won Chung, Jisang Ahn, Yizhuo Liang, Runyu Lu, Akshay Jajoo, Myungjin Lee, Mosharaf Chowdhury 4/17/2026

Cornfigurator: Automated Planning for Any-to-Any Multimodal Model Serving

Cornfigurator automates deployment planning for any-to-any multimodal models with heterogeneous computation paths and component scaling.

Ax Toshiaki Hori, Jonathan DeCastro, Deepak Gopinath, Avinash Balachandran, Guy Rosman 4/17/2026

Learning to Plan, Planning to Learn: Adaptive Hierarchical RL-MPC for Sample-Efficient Decision Making

Hierarchical approach combining reinforcement learning with MPC planning for sample-efficient decision making in structured planning problems.

Ax Suyan Dai, Gan Sun, Fazeng Li, Xu Tang, Qianqian Wang, Yang Cong 4/17/2026

Federated Multi-Task Clustering

Federated learning approach for spectral clustering in decentralized environments, capturing latent correlations across tasks.

Ax Waleed Khalid, Dmitry Ignatov, Radu Timofte 4/17/2026

From Memorization to Creativity: LLM as a Designer of Novel Neural Architectures

NNGPT framework uses LLMs for neural architecture synthesis through iterative supervised fine-tuning cycles generating validated PyTorch networks.

Ax Kun Liang, Clive Bai, Xin Xu, Chenming Tang, Sanwoo Lee, Weijie Liu, Saiyong Yang, Yunfang Wu 4/17/2026

ORBIT: On-policy Exploration-Exploitation for Controllable Multi-Budget Reasoning

ORBIT system for controlling reasoning budget in Large Reasoning Models via on-policy exploration-exploitation to reduce computational cost.

Ax Murat Bilgehan Ertan, Marten van Dijk 4/17/2026

Fundamental Limitations of Favorable Privacy-Utility Guarantees for DP-SGD

Theoretical analysis of differential privacy limitations in DP-SGD using f-differential privacy framework with shuffled sampling.

Ax Xingyue Huang, Xueying Ding, Mingxuan Ju, Yozen Liu, Neil Shah, Tong Zhao 4/17/2026

Threshold Differential Attention for Sink-Free, Ultra-Sparse, and Non-Dispersive Language Modeling

Threshold Differential Attention mechanism for long-context language modeling, addressing attention sink and sparsity issues without projection overhead.

Ax Badri N. Patro, Vijay S. Agneeswaran 4/17/2026

LLMOrbit: A Circular Taxonomy of Large Language Models -From Scaling Walls to Agentic AI Systems

Comprehensive circular taxonomy surveying 50+ LLMs from 2019-2025 across 15 organizations, documenting architectural innovations and scaling properties.

Ax Aditya Kumar, Mario A. Cypko, Oliver Amft 4/17/2026

Learning temporal embeddings from electronic health records of chronic kidney disease patients

Study of temporal embedding models for electronic health records of chronic kidney disease patients, balancing predictive performance with clinical interpretability.

Ax Rongzheng Wang, Yihong Huang, Muquan Li, Jiakai Li, Di Liang, Bob Simons, Pei Ke, Shuang Liang, Ke Qin 4/17/2026

Rethinking LLM-Driven Heuristic Design: Generating Efficient and Specialized Solvers via Dynamics-Aware Optimization

Framework for LLM-driven automated heuristic generation for combinatorial optimization with dynamics-aware evaluation beyond endpoint metrics.

Ax Wenlong Mou 4/17/2026

Continuous-time reinforcement learning: ellipticity enables model-free value function approximation

Research on model-free reinforcement learning for continuous-time processes with function approximation using ellipticity properties and Hilbert-space methods.

Ax Jianwen Chen, Xinyu Yang, Peng Xia, Arian Azarang, Yueh Z Lee, Gang Li, Hongtu Zhu, Yun Li, Beidi Chen, Huaxiu Yao 4/17/2026

MedVerse: Efficient and Reliable Medical Reasoning via DAG-Structured Parallel Execution

MedVerse framework enables parallel reasoning in LLMs for medical diagnosis using DAG-structured execution instead of sequential autoregressive decoding.

Ax Jonathan W. Siegel, Snir Hordan, Hannah Lawrence, Ali Syed, Nadav Dym 4/17/2026

Quantitative Approximation Rates for Group Equivariant Learning

arXiv paper on approximation rates for group equivariant neural networks. Theoretical ML research.

Ax Ganzhao Yuan 4/17/2026

OptEMA: Adaptive Exponential Moving Average for Stochastic Optimization with Zero-Noise Optimality

Adaptive optimizer using exponential moving average with theoretical guarantees for zero-noise convergence without restrictive boundedness conditions.

Ax Wenyue Hua, Sripad Karne, Qian Xie, Armaan Agrawal, Nikos Pagonas, Kostis Kaffes, Tianyi Peng 4/17/2026

AgentOpt v0.1 Technical Report: Client-Side Optimization for LLM-Based Agent

Client-side optimization framework for LLM-based agents that compose local tools, remote APIs, and multiple models efficiently.

Ax Wooin Lee, Hyun-Tae Kim 4/17/2026

SAGE: Sign-Adaptive Gradient for Memory-Efficient LLM Optimization

Memory-efficient optimizer for LLM training using sign-adaptive gradient updates to reduce optimizer state overhead.

Ax Pankayaraj Pathmanathan, Furong Huang 4/17/2026

Deliberative Alignment is Deep, but Uncertainty Remains: Inference time safety improvement in reasoning via attribution of unsafe behavior to base model

Studies deliberative alignment approach for improving LLM safety at inference time through reasoning model distillation and behavior attribution.

Ax Miit Daga, Swarna Priya Ramu 4/17/2026

Not All Forgetting Is Equal: Architecture-Dependent Retention Dynamics in Fine-Tuned Image Classifiers

Studies which samples are forgotten during fine-tuning of image classifiers and whether forgetting patterns depend on architecture.

Ax Denizalp Goktas, Gerardo Ria\~no-Brice\~no, Alif Abdullah, Aryan Nair, Chenkai Shen, Beatriz de Lucio, Alexandra Magnusson, Farhan Mashrur, Ahmed Abdulla, Shawrna Sen, Mahitha Thippireddy, Gregory Schwartz, Amy Greenwald 4/17/2026

TempusBench: An Evaluation Framework for Time-Series Forecasting

Comprehensive evaluation framework for assessing time-series foundation models with benchmarks and community standards.

Ax Hossem Eddine Hafidi, Elisabetta De Giovanni, Teodoro Montanaro, Ilaria Sergi, Massimo De Vittorio, Luigi Patrono 4/17/2026

Drowsiness-Aware Adaptive Autonomous Braking System based on Deep Reinforcement Learning for Enhanced Road Safety

Deep reinforcement learning system for autonomous braking that adapts to driver drowsiness using physiological state detection.

Ax Yuwei Yin, Giuseppe Carenini 4/17/2026

Improving Language Models with Intentional Analysis

Proposes intentional analysis framework to improve language model reasoning by explicitly understanding user intent behind questions.

Ax Zeguan Xiao, Yun Chen, Guanhua Chen, Ke Tang 4/17/2026

Towards Bridging the Reward-Generation Gap in Direct Alignment Algorithms

Identifies reward-generation gap in Direct Alignment Algorithms like DPO and SimPO for LLM alignment, proposes methods to bridge the discrepancy.

Ax Ido Guy, Daniel Haimovich, Fridolin Linder, Nastaran Okati, Lorenzo Perini, Niek Tax, Mark Tygert 4/17/2026

Measuring multi-calibration

Proposes scalar metrics for measuring multi-calibration of probabilistic predictions across multiple subpopulations.

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