Isolater - Feed

Ax Haoran Zhang, Dongjun Kim, Seohyeon Cha, Haris Vikalo 3/2/2026

FedRot-LoRA: Mitigating Rotational Misalignment in Federated LoRA

FedRot-LoRA addresses rotational misalignment in federated fine-tuning of LLMs on decentralized data to reduce aggregation error.

Ax Jielin Qiu, Jianguo Zhang, Zixiang Chen, Liangwei Yang, Ming Zhu, Juntao Tan, Haolin Chen, Wenting Zhao, Rithesh Murthy, Roshan Ram, Akshara Prabhakar, Shelby Heinecke, Caiming, Xiong, Silvio Savarese, Huan Wang 3/2/2026

AudioCapBench: Quick Evaluation on Audio Captioning across Sound, Music, and Speech

AudioCapBench: benchmark for evaluating audio captioning of multimodal LLMs across sound, music, speech with 1,000 samples and LLM-as-Judge evaluation.

Ax Haowen Zhu, Ning Yin, Xiaogen Zhou 3/2/2026

3D Modality-Aware Pre-training for Vision-Language Model in MRI Multi-organ Abnormality Detection

MedMAP pre-training framework for vision-language models on 3D MRI data with modality-specific alignment for multi-organ abnormality detection.

Ax Wei Luo, Yangfan Ou, Jin Deng, Zeshuai Deng, Xiquan Yan, Zhiquan Wen, Mingkui Tan 3/2/2026

ProtoDCS: Towards Robust and Efficient Open-Set Test-Time Adaptation for Vision-Language Models

ProtoDCS framework for test-time adaptation of vision-language models under distribution shift in open-set scenarios.

Ax Zitong Xu, Yuqing Wu, Yue Zhao 3/2/2026

TRIZ-RAGNER: A Retrieval-Augmented Large Language Model for TRIZ-Aware Named Entity Recognition in Patent-Based Contradiction Mining

TRIZ-RAGNER applies retrieval-augmented LLMs for named entity recognition in patent contradiction mining for systematic innovation.

Ax Yongzhong Xu 3/2/2026

Optimizer-Induced Low-Dimensional Drift and Transverse Dynamics in Transformer Training

Analysis of transformer training geometry showing parameter updates organize into dominant drift direction with oscillatory transverse dynamics.

Ax Wenzhe Zhao, Yang Zhao, Ganchao Liu, Zhiyu Jiang, Dandan Ma, Zihao Li, Xuelong Li 3/2/2026

SAGE-LLM: Towards Safe and Generalizable LLM Controller with Fuzzy-CBF Verification and Graph-Structured Knowledge Retrieval for UAV Decision

SAGE-LLM architecture combining LLMs with formal safety verification (Fuzzy-CBF) and graph-structured knowledge for safe UAV autonomous decision-making.

Ax Hariz Yet, Nguyen Thanh Tam, Mao V. Ngo, Lim Yi Shen, Lin Wei, Jihong Park, Binbin Chen, Tony Q. S. Quek 3/2/2026

SLA-Aware Distributed LLM Inference Across Device-RAN-Cloud

System design for distributed LLM inference across device, RAN-edge, and cloud tiers with latency constraints for 5G embodied AI applications.

Ax Seungdong Yoa, Sanghyu Yoon, Suhee Yoon, Dongmin Kim, Ye Seul Sim, Junhyun Lee, Woohyung Lim 3/2/2026

From Static Benchmarks to Dynamic Protocol: Agent-Centric Text Anomaly Detection for Evaluating LLM Reasoning

Agent-centric benchmarking paradigm where autonomous agents dynamically generate, validate, and solve problems to evaluate LLM reasoning capabilities.

Ax Hanping Zhang, Yuhong Guo 3/2/2026

Bridging Dynamics Gaps via Diffusion Schr\"odinger Bridge for Cross-Domain Reinforcement Learning

BDGxRL uses Diffusion Schrödinger Bridge to enable cross-domain reinforcement learning policy transfer when target domain interaction is unavailable.

Ax Aleksandr Ananikian (Saint-Petersburg University), Daniil Drozdov (Saint-Petersburg University), Konstantin Yakovlev (Saint-Petersburg University) 3/2/2026

UPath: Universal Planner Across Topological Heterogeneity For Grid-Based Pathfinding

UPath proposes learning-based heuristics for A* pathfinding on grid maps using deep neural networks across heterogeneous topologies.

Ax Tiantong Wang, Xinyu Yan, Tiantong Wu, Yurong Hao, Yong Jiang, Fei Huang, Wei Yang Bryan Lim 3/2/2026

MPU: Towards Secure and Privacy-Preserving Knowledge Unlearning for Large Language Models

MPU framework for privacy-preserving machine unlearning in LLMs without sharing server parameters or client forget sets.

Ax Tadahisa Okuda, Shohei Shimizu, Thong Pham, Tatsuyoshi Ikenoue, Shingo Fukuma 3/2/2026

Operationalizing Longitudinal Causal Discovery Under Real-World Workflow Constraints

Research on applying causal discovery algorithms to real-world longitudinal data with institutional workflow constraints.

Ax Tianci Tang, Tielong Cai, Hongwei Wang, Gaoang Wang 3/2/2026

See, Act, Adapt: Active Perception for Unsupervised Cross-Domain Visual Adaptation via Personalized VLM-Guided Agent

Sea² framework using vision-language models and autonomous agents for unsupervised cross-domain visual adaptation without fine-tuning on target data.

Ax Xiang Li, Nan Jiang, Yuheng Zhang 3/2/2026

Beyond State-Wise Mirror Descent: Offline Policy Optimization with Parameteric Policies

Theoretical research on offline reinforcement learning with general function approximation and parametric policies, extending beyond finite action spaces.

Ax George Papadopoulos, George A. Vouros 3/2/2026

Learning to maintain safety through expert demonstrations in settings with unknown constraints: A Q-learning perspective

Q-learning approach for learning safe policies from expert demonstrations with unknown constraints and non-observable costs.

Ax Junkang Liu, Fanhua Shang, Yuxuan Tian, Hongying Liu, Yuanyuan Liu 3/2/2026

FedNSAM:Consistency of Local and Global Flatness for Federated Learning

Federated learning optimization achieving consistency between local and global model flatness under data heterogeneity.

Ax Xuhui Dou, Hayretdin Bahsi, Alejandro Guerra-Manzanares 3/2/2026

Enhancing Continual Learning for Software Vulnerability Prediction: Addressing Catastrophic Forgetting via Hybrid-Confidence-Aware Selective Replay for Temporal LLM Fine-Tuning

Continual fine-tuning method for LLM-based vulnerability detection addressing catastrophic forgetting under temporal distribution shift.

Ax Wei Lian, Alejandro Guerra-Manzanares 3/2/2026

MI$^2$DAS: A Multi-Layer Intrusion Detection Framework with Incremental Learning for Securing Industrial IoT Networks

Multi-layer intrusion detection framework with incremental learning for Industrial IoT networks handling novel cyber threats.

Ax Qihua Dong, Kuo Yang, Lin Ju, Handong Zhao, Yitian Zhang, Yizhou Wang, Huimin Zeng, Jianglin Lu, Yun Fu 3/2/2026

Ref-Adv: Exploring MLLM Visual Reasoning in Referring Expression Tasks

Adversarial benchmark testing MLLM visual reasoning and grounding capabilities in referring expression comprehension tasks.

Ax Pramit Saha, Mohammad Alsharid, Joshua Strong, J. Alison Noble 3/2/2026

Experience-Guided Self-Adaptive Cascaded Agents for Breast Cancer Screening and Diagnosis with Reduced Biopsy Referrals

Multi-agent cascaded framework for breast ultrasound screening reducing unnecessary biopsy referrals through selective decision-making.

Ax Jiaqi Tang, Shaoyang Zhang, Xiaoqi Wang, Jiaying Zhou, Yang Liu, Qingchao Chen 3/2/2026

The Geometry of Transfer: Unlocking Medical Vision Manifolds for Training-Free Model Ranking

Transferability estimation method for selecting optimal medical foundation models for segmentation without retraining.

Ax Dongxu Zhang, Yiding Sun, Pengcheng Li, Yumou Liu, Hongqiang Lin, Haoran Xu, Xiaoxuan Mu, Liang Lin, Wenbiao Yan, Ning Yang, Chaowei Fang, Juanjuan Zhao, Jihua Zhu, Conghui He, Cheng Tan 3/2/2026

PointCoT: A Multi-modal Benchmark for Explicit 3D Geometric Reasoning

Multimodal benchmark evaluating MLLMs on explicit 3D geometric reasoning with point clouds, exposing geometric hallucinations.

Ax Oscar Hill, Mateo Espinosa Zarlenga, Mateja Jamnik 3/2/2026

Hierarchical Concept-based Interpretable Models

Hierarchical concept embedding models improving neural network interpretability through human-readable concept representations.

Ax Guy Hadad, Shadi Iskander, Oren Kalinsky, Sofia Tolmach, Ran Levy, Haggai Roitman 3/2/2026

HotelQuEST: Balancing Quality and Efficiency in Agentic Search

Benchmark for agentic search systems balancing quality and efficiency, addressing underspecified user preferences in LLM-powered retrieval.

Ax Magda Dubois, Cozmin Ududec, Christopher Summerfield, Lennart Luettgau 3/2/2026

Ask don't tell: Reducing sycophancy in large language models

Controlled experimental studies identifying what triggers and prevents sycophancy in LLMs, an alignment failure in advisory contexts.

Ax Florent Delgrange 3/2/2026

Foundation World Models for Agents that Learn, Verify, and Adapt Reliably Beyond Static Environments

Vision for foundation world models enabling autonomous agents to learn, verify, and adapt reliably in open, non-static environments.

Ax Zhicheng Fang, Jingjie Zheng, Chenxu Fu, Wei Xu 3/2/2026

Jailbreak Foundry: From Papers to Runnable Attacks for Reproducible Benchmarking

Multi-agent workflow system that translates jailbreak papers into executable modules for unified benchmarking of LLM robustness techniques.

Ax Na Min An, Yoonna Jang, Yusuke Hirota, Ryo Hachiuma, Isabelle Augenstein, Hyunjung Shim 3/2/2026

Interpretable Debiasing of Vision-Language Models for Social Fairness

Model-agnostic interpretable debiasing method for vision-language models to mitigate unintended social bias in black-box reasoning processes.

Ax Daniel Yang, Samuel Stante, Florian Redhardt, Lena Libon, Parnian Kassraie, Ido Hakimi, Barna P\'asztor, Andreas Krause 3/2/2026

RewardUQ: A Unified Framework for Uncertainty-Aware Reward Models

RewardUQ: Framework for uncertainty quantification in reward models used to align LLMs with human preferences, reducing annotation costs.

Ax Ferran Agullo, Joan Oliveras, Chen Wang, Alberto Gutierrez-Torre, Olivier Tardieu, Alaa Youssef, Jordi Torres, Josep Ll. Berral 3/2/2026

Data Driven Optimization of GPU efficiency for Distributed LLM Adapter Serving

Data-driven optimization pipeline for scheduling and caching hundreds of LLM adapters in distributed serving to maximize GPU throughput.

Ax Chenwei Jia, Baoting Li, Xuchong Zhang, Mingzhuo Wei, Bochen Lin, Hongbin Sun 3/2/2026

Quant Experts: Token-aware Adaptive Error Reconstruction with Mixture of Experts for Large Vision-Language Models Quantization

Quant Experts: Post-training quantization method for vision-language models using mixture of experts for token-aware adaptive error reconstruction.

Ax Donghao Huang, Zhaoxia Wang 3/2/2026

Task Complexity Matters: An Empirical Study of Reasoning in LLMs for Sentiment Analysis

Empirical study evaluating whether reasoning capabilities universally improve LLM performance across sentiment analysis tasks of varying complexity.

Ax Viet Bac Nguyen, Phuong Thai Nguyen 3/2/2026

Adaptive Correlation-Weighted Intrinsic Rewards for Reinforcement Learning

ACWI framework for dynamically balancing intrinsic and extrinsic rewards in sparse reward reinforcement learning through adaptive scaling.

Ax Jaekyung Cho 3/2/2026

Preference Packing: Efficient Preference Optimization for Large Language Models

Preference packing technique for efficient batch training of LLMs during preference optimization (RLHF), improving resource utilization.

Ax Sara Nabhani, Federico Pianzola, Khalid Al-Khatib, Malvina Nissim 3/2/2026

ARGUS: Seeing the Influence of Narrative Features on Persuasion in Argumentative Texts

ARGUS framework studying how narrative features in argumentative texts influence persuasion using corpus analysis and modeling.

Ax Vikash Singh, Debargha Ganguly, Haotian Yu, Chengwei Zhou, Prerna Singh, Brandon Lee, Vipin Chaudhary, Gourav Datta 3/2/2026

Toward Guarantees for Clinical Reasoning in Vision Language Models via Formal Verification

Formal verification approach for vision-language models drafting radiology reports to ensure logical consistency in clinical reasoning.

Ax James L. Zainaldin, Cameron Pattison, Manuela Marai, Jacob Wu, Mark J. Schiefsky 3/2/2026

Terminology Rarity Predicts Catastrophic Failure in LLM Translation of Low-Resource Ancient Languages: Evidence from Ancient Greek

Systematic evaluation of LLM machine translation for Ancient Greek technical prose, showing terminology rarity predicts translation failures.

Ax Yuxuan Liu, Weikai Xu, Kun Huang, Changyu Chen, Jiankun Zhao, Pengzhi Gao, Wei Liu, Jian Luan, Shuo Shang, Bo Du, Ji-Rong Wen, Rui Yan 3/2/2026