Isolater - Feed

Ax Mahdi Alizadeh 4/13/2026

eBandit: Kernel-Driven Reinforcement Learning for Adaptive Video Streaming

eBandit uses eBPF and multi-armed bandit reinforcement learning in Linux kernel for adaptive video bitrate selection with improved network signal visibility.

Ax Sophie Wu, Andrew Piper 4/13/2026

Lessons Without Borders? Evaluating Cultural Alignment of LLMs Using Multilingual Story Moral Generation

Evaluates cultural alignment of LLMs across 14 language-culture pairs using multilingual story moral generation task and dataset.

Ax Marc B\"ohlen, Sai Krishna 4/13/2026

Scrapyard AI

Investigates opportunities for resource-constrained AI research using obsolete yet capable discarded models from AI production cycles.

Ax Chris Hicks, Elizabeth Bates, Shae McFadden, Isaac Symes Thompson, Myles Foley, Ed Chapman, Nickolas Espinosa Dice, Ankita Samaddar, Joshua Sylvester, Himanshu Neema, Nicholas Butts, Nate Foster, Ahmad Ridley, Zoe M, Paul Jones 4/13/2026

Building Better Environments for Autonomous Cyber Defence

Workshop report on designing reinforcement learning environments for autonomous cyber defense applications.

Ax Fatih Cagatay Akyon, Alptekin Temizel 4/13/2026

SenBen: Sensitive Scene Graphs for Explainable Content Moderation

SenBen large-scale scene graph benchmark for explainable content moderation with visual grounding and sensitivity annotations.

Ax Mehran Taghian, Yunke Peng, Xing Huang, Yao Wang, Yaoyuan Wang, Wei Guo, Yuanyong Luo, Tianchi Hu, Junsong Wang, Xin Wang, Hu Liu, Yu Cheng, Ziwei Yu, Hongliang Li, Mehdi Rahimifar, Lei Yan, Xuefei Wang, Zhuang Ma, Lei Liu, Hui Yu, Anandharaju Durai Raju, Hoang Le, Hei Yi Mak, Tanzila Rahman, Shadan Golestan 4/13/2026

HiFloat4 Format for Language Model Pre-training on Ascend NPUs

HiFloat4 low-precision floating-point format for efficient 4-bit LLM pre-training on Ascend NPU hardware.

Ax Jinqi Luo, Jinyu Yang, Tal Neiman, Lei Fan, Bing Yin, Son Tran, Mubarak Shah, Ren\'e Vidal 4/13/2026

Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs

Dictionary-aligned concept control method for safeguarding multimodal LLMs against malicious queries at inference time.

Ax Cyrus Zhou, Yufei Jin, Yilin Xu, Yu-Chiang Wang, Chieh-Ju Chao, Monica S. Lam 4/13/2026

Scalable High-Recall Constraint-Satisfaction-Based Information Retrieval for Clinical Trials Matching

Constraint-satisfaction-based retrieval system for matching patient profiles to clinical trials with high recall and precision.

Ax Greg Nyilasy, Brock Bastian, Jennifer Overbeck, Abraham Ryan Ade Putra Hito 4/13/2026

AI-Induced Human Responsibility (AIHR) in AI-Human teams

Empirical study on how humans allocate responsibility in AI-human hybrid workflows using AI-assisted lending experiments.

Ax Mintong Kang, Chen Fang, Bo Li 4/13/2026

AudioGuard: Toward Comprehensive Audio Safety Protection Across Diverse Threat Models

AudioGuard framework for comprehensive audio safety protection including voice impersonation, speaker attributes, and compositional harms.

Ax Tokio Kajitsuka, Ukyo Honda, Sho Takase 4/13/2026

Revisiting the Capacity Gap in Chain-of-Thought Distillation from a Practical Perspective

Re-examines capacity gap in chain-of-thought distillation, finding student models often outperform teacher distillation baselines.

Ax Chengjie Fan, Cong Pan, Zijian Liu, Ningzhong Liu, Jie Qin 4/13/2026

HTNav: A Hybrid Navigation Framework with Tiered Structure for Urban Aerial Vision-and-Language Navigation

HTNav framework for aerial vision-and-language navigation combining visual perception with language instructions in urban environments.

Ax Xinyu Zhang, Zurong Mai, Qingmei Li, Zjin Liao, Yibin Wen, Yuhang Chen, Xiaoya Fan, Chan Tsz Ho, Bi Tianyuan, Haoyuan Liang, Ruifeng Su, Zihao Qian, Juepeng Zheng, Jianxi Huang, Yutong Lu, Haohuan Fu 4/13/2026

HM-Bench: A Comprehensive Benchmark for Multimodal Large Language Models in Hyperspectral Remote Sensing

HM-Bench benchmark evaluates multimodal LLMs on hyperspectral remote sensing image understanding tasks.

Ax Hengran Zhang, Minghao Tang, Keping Bi, Jiafeng Guo 4/13/2026

Beyond Relevance: Utility-Centric Retrieval in the LLM Era

RAG systems should optimize for utility (task completion) rather than topical relevance when retrieving documents for LLMs.

Ax Rares-Alexandru Roscan, Gabriel Petre1, Adrian-Marius Dumitran, Angela-Liliana Dumitran 4/13/2026

MuTSE: A Human-in-the-Loop Multi-use Text Simplification Evaluator

MuTSE: Human-in-the-loop evaluator tool for systematically comparing LLM text simplification outputs across different prompting strategies and architectures.

Ax Mintae Kim, Koushil Sreenath 4/13/2026

WOMBET: World Model-based Experience Transfer for Robust and Sample-efficient Reinforcement Learning

WOMBET: Framework for reinforcement learning that generates and transfers experience data between source and target robotic tasks for sample efficiency.

Ax Keyu Li, Jin Gao, Dequan Wang 4/13/2026

Aligned Agents, Biased Swarm: Measuring Bias Amplification in Multi-Agent Systems

Aligned Agents, Biased Swarm: Empirical study measuring how multi-agent system topologies and feedback loops amplify bias in emergent behaviors.

Ax Avni Mittal, Shanu Kumar, Sandipan Dandapat, Monojit Choudhury 4/13/2026

Litmus (Re)Agent: A Benchmark and Agentic System for Predictive Evaluation of Multilingual Models

Litmus ReAgent: Benchmark and agentic system for evaluating multilingual LLM performance prediction across 1,500 questions spanning six tasks and five evidence scenarios.

Ax Jihwan Oh, Soowon Oh, Murad Aghazada, Minchan Jeong, Sungnyun Kim, Se-Young Yun 4/13/2026

PerMix-RLVR: Preserving Persona Expressivity under Verifiable-Reward Alignment

PerMix-RLVR: Training method for aligning LLM personas with reward models while preserving output diversity, avoiding inference-time computation overhead.

Ax Zhiyu Zhou, Peilin Liu, Ruoxuan Zhang, Luyang Zhang, Cheng Zhang, Hongxia Xie, Wen-Huang Cheng 4/13/2026

PinpointQA: A Dataset and Benchmark for Small Object-Centric Spatial Understanding in Indoor Videos

PinpointQA dataset and benchmark for evaluating small object localization and spatial reasoning in video MLLMs.

Ax Xiaoke Guo, Songze Li, Zhiqiang Liu, Zhaoyan Gong, Yuanxiang Liu, Huajun Chen, Wen Zhang 4/13/2026

ASTRA: Adaptive Semantic Tree Reasoning Architecture for Complex Table Question Answering

ASTRA: adaptive semantic tree reasoning architecture for LLM-based complex table question answering.

Ax Andre Bacellar 4/13/2026

Regime-Conditional Retrieval: Theory and a Transferable Router for Two-Hop QA

Regime-conditional retrieval with transferable router for two-hop question answering with theoretical foundations.

Ax Qixuan Huang, Khalid Zaman, Masashi Unoki 4/13/2026

Noise-Aware In-Context Learning for Hallucination Mitigation in ALLMs

Noise-aware in-context learning approach to mitigate hallucinations in auditory large language models.

Ax Zedian Shao, Hongbin Liu, Yuepeng Hu, Neil Zhenqiang Gong 4/13/2026

Leave My Images Alone: Preventing Multi-Modal Large Language Models from Analyzing Images via Visual Prompt Injection

ImageProtector prevents multi-modal LLMs from analyzing images via visual prompt injection attacks.

Ax Chenjie Yang, Yutian Jiang, Chenyu Wu 4/13/2026

Skill-Conditioned Visual Geolocation for Vision-Language

Vision-language models for image geolocation with structured geographic reasoning and autonomous self-evolution.

Ax Yeonjun Hwang, Sungyong Park, Minju Kim, Dongha Lee, Jinyoung Yeo 4/13/2026

CONDESION-BENCH: Conditional Decision-Making of Large Language Models in Compositional Action Space

CONDESION-BENCH evaluates LLM decision-making with compositional action spaces and conditional feasibility constraints.

Ax Mauricio Fadel Argerich, Jonathan F\"urst, Marta Pati\~no-Mart\'inez 4/13/2026

Watt Counts: Energy-Aware Benchmark for Sustainable LLM Inference on Heterogeneous GPU Architectures

Watt Counts: open-access energy consumption benchmark for LLM inference across 50 models and 10 GPU architectures.

Ax Min Young Baeg, Yoon-Yeong Kim 4/13/2026

PDE-regularized Dynamics-informed Diffusion with Uncertainty-aware Filtering for Long-Horizon Dynamics

PDYffusion combines diffusion models with physics-informed dynamics for long-horizon spatiotemporal prediction.

Ax Guoqing Wang, Pin Tang, Xiangxuan Ren, Guodongfang Zhao, Bailan Feng, Chao Ma 4/13/2026

Learning Vision-Language-Action World Models for Autonomous Driving

Vision-Language-Action models for autonomous driving combining perception, reasoning, and temporal dynamics modeling.

Ax Harry Proshian, Nikita Severin, Sergey Nikolenko, Kireev Ivan, Andrey Savchenko, Ivan Sergeev, Maria Postnova, Ilya Makarov 4/13/2026

Beyond Isolated Clients: Integrating Graph-Based Embeddings into Event Sequence Models

Method integrating graph-based embeddings into event sequence models for improved user prediction on digital platforms.

Ax Li Huang, Zhongxin Liu, Yifan Wu, Tao Yin, Dong Li, Jichao Bi, Nankun Mu, Hongyu Zhang, Meng Yan 4/13/2026

DeepGuard: Secure Code Generation via Multi-Layer Semantic Aggregation

DeepGuard improves secure code generation by LLMs through multi-layer semantic aggregation to mitigate vulnerable patterns.

Ax Akshit Jindal, Saket Anand, Chetan Arora, Vikram Goyal 4/13/2026

CLIP-Inspector: Model-Level Backdoor Detection for Prompt-Tuned CLIP via OOD Trigger Inversion

CLIP-Inspector detects backdoor attacks in prompt-tuned vision-language models through out-of-distribution trigger inversion.

Ax Tommy Shaffer Shane, Simon Mylius, Hamish Hobbs 4/13/2026

Scheming in the wild: detecting real-world AI scheming incidents with open-source intelligence

Research on detecting covert misaligned AI behavior in real-world settings using open-source intelligence methods.

Ax Chenhao Ye, Huaizheng Zhang, Mingcong Han, Baoquan Zhong, Xiang Li, Qixiang Chen, Xinyi Zhang, Weidong Zhang, Kaihua Jiang, Wang Zhang, He Sun, Wencong Xiao, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau 4/13/2026

TensorHub: Scalable and Elastic Weight Transfer for LLM RL Training

TensorHub introduces Reference-Oriented Storage for efficient weight transfer in LLM reinforcement learning across heterogeneous computational resources.

Ax Changi Hong, Yoonah Song, Hwayoung Park, Chaewoon Bang, Dayeon Gu, Do Hyun Lee, Hong Kook Kim 4/13/2026

PS-TTS: Phonetic Synchronization in Text-to-Speech for Achieving Natural Automated Dubbing

PS-TTS method for phonetic synchronization in automated dubbing, addressing duration and lip-sync challenges in AI-based video translation.

Ax Peng Wang (X-LANCE Lab, Shanghai Jiao Tong University), Yanqiao Zhu (X-LANCE Lab, Shanghai Jiao Tong University), Zixuan Jiang (X-LANCE Lab, Shanghai Jiao Tong University), Qinyuan Chen (School of Computer Science, Fudan University), Xingjian Zhao (School of Computer Science, Fudan University), Xipeng Qiu (School of Computer Science, Fudan University), Wupeng Wang (Tongyi Fun Team, Alibaba Group), Zhifu Gao (Tongyi Fun Team, Alibaba Group), Xiangang Li (Tongyi Fun Team, Alibaba Group), Kai Yu (X-LANCE Lab, Shanghai Jiao Tong University), Xie Chen (X-LANCE Lab, Shanghai Jiao Tong University) 4/13/2026