Isolater - Feed

Ax Jianfei Xiao, Xiang Yu, Chengbing Wang, Wuqiang Zheng, Xinyu Lin, Kaining Liu, Hongxun Ding, Yang Zhang, Wenjie Wang, Fuli Feng, Xiangnan He 3/31/2026

AlpsBench: An LLM Personalization Benchmark for Real-Dialogue Memorization and Preference Alignment

AlpsBench: evaluation benchmark for LLM personalization using real dialogue data covering memorization and preference alignment tasks.

Ax Seonok Kim 3/31/2026

LITTA: Late-Interaction and Test-Time Alignment for Visually-Grounded Multimodal Retrieval

LITTA framework for multimodal document retrieval using query expansion to improve evidence page retrieval from visually rich documents without retraining.

Ax Qiming Zhu, Zhirui Fang, Tianming Zhang, Chuanxiu Liu, Xiaoke Jiang, Lei Zhang 3/31/2026

SpatialPoint: Spatial-aware Point Prediction for Embodied Localization

SpatialPoint framework for embodied AI predicting executable 3D points from visual-language input, enabling physical interaction and navigation in environments.

Ax Tianyu Wu, Lin Zhu 3/31/2026

PI-Mamba: Linear-Time Protein Backbone Generation via Spectrally Initialized Flow Matching

Physics-Informed Mamba generative model for protein backbone design using spectral initialization and flow matching for efficient long-sequence generation.

Ax Netanel Eliav (Machine Human Intelligence Lab) 3/31/2026

The Cognitive Divergence: AI Context Windows, Human Attention Decline, and the Delegation Feedback Loop

Theoretical analysis of divergence between expanding LLM context windows (512 to 2M tokens) and contracting human attention capacity from 2017-2026.

Ax Kamer Ali Yuksel, Abdul Basit Anees, Ashraf Elneima, Sanjika Hewavitharana, Mohamed Al-Badrashiny, Hassan Sawaf 3/31/2026

Agentic AI for Human Resources: LLM-Driven Candidate Assessment

Modular LLM-driven framework for automated candidate assessment in recruitment, integrating job descriptions, CVs, and interviews for structured evaluation reports.

Ax Andres Alonso-Robisco, Carlos Esparcia, Francisco Jare\~no 3/31/2026

On the Carbon Footprint of Economic Research in the Age of Generative AI

Study of carbon footprint in economic research workflows using generative AI as tool, shifting analysis from model-level to workflow-level impact assessment.

Ax Marcin Abram 3/31/2026

Toward Evaluation Frameworks for Multi-Agent Scientific AI Systems

Analysis of benchmarking challenges for multi-agent scientific AI systems including contamination risks, ground truth issues, and tool use complications.

Ax Huanrong Liu, Chunlin Tian, Tongyu Jia, Tailai Zhou, Qin Liu, Yu Gao, Yutong Ban, Yun Gu, Guy Rosman, Xin Ma, Qingbiao Li 3/31/2026

SutureAgent: Learning Surgical Trajectories via Goal-conditioned Offline RL in Pixel Space

Goal-conditioned offline RL agent for learning surgical needle trajectories from endoscopic video, enabling robot-assisted suturing with trajectory prediction.

Ax Cosimo Spera 3/31/2026

Capability Safety as Datalog: A Foundational Equivalence

Theoretical proof that capability safety is exactly representable as propositional Datalog, enabling algorithmic optimization and incremental maintenance.

Ax Zecheng Zhang, Han Zheng, Yue Xu 3/31/2026

SEAR: Schema-Based Evaluation and Routing for LLM Gateways

Schema-based evaluation and routing system for multi-provider LLM gateways enabling fine-grained quality assessment and intelligent request routing across models.

Ax Martina G. Vilas, Timothy Schauml\"offel, Gemma Roig 3/31/2026

Contextual inference from single objects in Vision-Language models

Systematic analysis of how vision-language models extract scene context from single objects, examining robustness implications through behavioral and mechanistic study.

Ax Qinghui Chen, Zekai Zhang, Zaigui Zhang, Kai Zhang, Dagang Li, Wenmin Wang, Jinglin Zhang, Cong Liu 3/31/2026

Distilled Large Language Model-Driven Dynamic Sparse Expert Activation Mechanism

Distilled LLM-driven Sparse Mixture-of-Experts framework for visual recognition combining text guidance with dynamic expert activation for improved generalization.

Ax Mariana D\'oria Prata Lima, Gilson Antonio Giraldi, Jaime S. Cardoso 3/31/2026

Ordinal Semantic Segmentation Applied to Medical and Odontological Images

Semantic segmentation method incorporating ordinal relationships between classes for improved accuracy on medical and odontological images.

Ax Guangfu Guo, Xiaoqian Lu, Yue Feng, Mingming Sun 3/31/2026

Beyond Static Visual Tokens: Structured Sequential Visual Chain-of-Thought Reasoning

Structured Sequential Visual Chain-of-Thought reasoning for multimodal LLMs, enabling selective sequential visual attention instead of static visual tokens.

Ax Guifeng Deng, Pan Wang, Jiquan Wang, Shuying Rao, Junyi Xie, Wanjun Guo, Tao Li, Haiteng Jiang 3/31/2026

SleepVLM: Explainable and Rule-Grounded Sleep Staging via a Vision-Language Model

Vision-language model for sleep staging from polysomnography waveforms, generating clinician-readable explanations grounded in medical scoring criteria.

Ax Yifei Dong, Fengyi Wu, Yilong Dai, Lingdong Kong, Guangyu Chen, Xu Zhu, Qiyu Hu, Tianyu Wang, Johnalbert Garnica, Feng Liu, Siyu Huang, Qi Dai, Zhi-Qi Cheng 3/31/2026

Language-Conditioned World Modeling for Visual Navigation

Language-conditioned visual navigation using world modeling, where embodied agents follow natural language instructions based on visual observations and continuous control.

Ax Yousung Lee, Dongsoo Har 3/31/2026

Steering Sparse Autoencoder Latents to Control Dynamic Head Pruning in Vision Transformers (Student Abstract)

Framework integrating Sparse Autoencoders with dynamic head pruning in Vision Transformers for interpretable and controllable model compression.

Ax Yassine Bougacha (DGA), Geoffrey Delhomme (DGA), M\'elanie Ducoffe (DGA), Augustin Fuchs (DGA), Jean-Brice Ginestet (DGA), Jacques Girard, Sofiane Kraiem, Franck Mamalet, Vincent Mussot, Claire Pagetti, Thierry Sammour 3/31/2026

LARD 2.0: Enhanced Datasets and Benchmarking for Autonomous Landing Systems

Enhanced datasets and benchmarking framework for autonomous landing systems, addressing object detection dataset limitations through diverse data sources.

Ax Jian Guan, Huolong Wu, Zhenzhong Wang, Gary G. Yen, Min Jiang 3/31/2026

Training-Free Diffusion-Driven Modeling of Pareto Set Evolution for Dynamic Multiobjective Optimization

Diffusion-based approach for modeling Pareto set evolution in dynamic multiobjective optimization problems without requiring training.

Ax David Brundage 3/31/2026

Generating Synthetic Wildlife Health Data from Camera Trap Imagery: A Pipeline for Alopecia and Body Condition Training Data

Pipeline for generating synthetic training data from camera trap imagery to enable ML models for automated wildlife health screening (alopecia and body condition detection).

Ax Shakil Mia, Umme Habiba, Urmi Akter, SK Rezwana Quadir Raisa, Jeba Maliha, Md. Iqbal Hossain, Md. Shakhauat Hossan Sumon 3/31/2026

Tiny-ViT: A Compact Vision Transformer for Efficient and Explainable Potato Leaf Disease Classification

Compact Vision Transformer for potato leaf disease classification using efficient architecture design and explainability features.

Ax Chen Zheng, Yuxuan Lai, Haoyang Lu, Wentao Ma, Jitao Yang, Jian Wang 3/31/2026

Aesthetic Assessment of Chinese Handwritings Based on Vision Language Models

Application of vision-language models to assess handwritten Chinese character quality and generate actionable feedback for learners.

Ax Mehmet Kaan Erol 3/31/2026

Edge Reliability Gap in Vision-Language Models: Quantifying Failure Modes of Compressed VLMs Under Visual Corruption

Study comparing failure modes of compressed vision-language models on edge devices, proposing taxonomy of error types (Object Blindness, Semantic Drift, Prior Bias) across different model sizes.

Ax Paolo Cupini, Francesco Pierri 3/31/2026

From Content to Audience: A Multimodal Annotation Framework for Broadcast Television Analytics

Framework for automated semantic annotation of broadcast television using multimodal LLMs, evaluating different pipeline architectures and input configurations for video understanding.

Ax Eugene Lee, Yu-Chi Lin, Jiajie Diao 3/31/2026

Learning to Select Visual In-Context Demonstrations

Method to improve in-context demonstration selection for multimodal LLMs by reformulating it as a sequential task rather than relying on k-NN similarity, improving performance on visual regression tasks.

Ax Shuozhi Yuan, Jinqing Wang, Zihao Liu, Miaomiao Yuan, Haoran Peng, Jin Zhao, Bingwen Wang, Haoyi Wang 3/31/2026

TED: Training-Free Experience Distillation for Multimodal Reasoning

TED proposes a training-free knowledge distillation framework for multimodal models using context-based approaches instead of parameter optimization, enabling deployment in resource-constrained environments.

Ax Sergio Y. Hayashi, Nina S. T. Hirata 3/31/2026

Limits of Imagery Reasoning in Frontier LLM Models

arXiv research on spatial reasoning gaps in frontier LLMs, testing external imagery module as cognitive prosthetic for mental rotation tasks.

Ax Baochen Xiong, Yifan Xu, Xiaoshan Yang, Yaguang Song, Yaowei Wang, Changsheng Xu 3/31/2026

A Step Toward Federated Pretraining of Multimodal Large Language Models

Federated learning approach for pre-training multimodal large language models on distributed privacy-sensitive data without centralized data collection.

Ax Hannah Collison, Benjamin Van Durme, Daniel Khashabi 3/31/2026

CRISP: Characterizing Relative Impact of Scholarly Publications

CRISP: LLM-based method to jointly rank cited papers within citing papers by analyzing citation context, mitigating positional bias through randomized ranking.

Ax Hayder Saad Abdulbaqi, Mohammed Hadi Rahim, Mohammed Hassan Hadi, Haider Ali Aboud, Ali Hussein Allawi 3/31/2026

PhyDCM: A Reproducible Open-Source Framework for AI-Assisted Brain Tumor Classification from Multi-Sequence MRI

Open-source reproducible framework for deep learning-based brain tumor classification from multi-sequence MRI with emphasis on accessibility and academic development.

Ax Jelena Markovic-Voronov, Kayhan Behdin, Yuanda Xu, Zhengze Zhou, Zhipeng Wang, Rahul Mazumder 3/31/2026

Robust Batch-Level Query Routing for Large Language Models under Cost and Capacity Constraints

Batch-level routing framework for LLM queries that optimizes model assignment under cost, GPU capacity, and concurrency constraints with resource-aware optimization.

Ax Gesina Schwalbe, Mert Keser, Moritz Bayerkuhnlein, Edgar Heinert, Annika M\"utze, Marvin Keller, Sparsh Tiwari, Georgii Mikriukov, Diedrich Wolter, Jae Hee Lee, Matthias Rottmann 3/31/2026

Explaining, Verifying, and Aligning Semantic Hierarchies in Vision-Language Model Embeddings

Framework to interpret and verify semantic hierarchies in CLIP vision-language model embeddings through agglomerative clustering and hierarchy alignment.

Ax Huanshuo Dong, Hao Wu, Hong Wang, Qin-Yi Zhang, Zhezheng Hao 3/31/2026

DSO: Dual-Scale Neural Operators for Stable Long-term Fluid Dynamics Forecasting

Neural operators with dual-scale architecture for stable long-term fluid dynamics forecasting by addressing local detail blurring and precision issues in PDE modeling.

Ax Filippo Cenacchi 3/31/2026

Sparse-by-Design Cross-Modality Prediction: L0-Gated Representations for Reliable and Efficient Learning

Unified sparsification framework for cross-modality prediction across graphs, language, and tabular data using L0-gated representations for efficient multimodal learning.

Ax Jin Chen, Yifeng Lin, Chao Zeng, Si Wu, Tiesong Zhao 3/31/2026

The Language of Touch: Translating Vibrations into Text with Dual-Branch Learning

First approach to vibrotactile captioning, generating natural language descriptions from vibration signals for VR and human-computer interaction applications.

Ax Xinyi Duan, Yuanrong Tang, Jiangtao Gong 3/31/2026

GroupRAG: Cognitively Inspired Group-Aware Retrieval and Reasoning via Knowledge-Driven Problem Structuring

GroupRAG framework combining retrieval-augmented generation with cognitively-inspired group-aware reasoning and knowledge-driven problem structuring.

Ax Zhiyuan Cheng, Longying Lai, Yue Liu 3/31/2026

Resolving the Robustness-Precision Trade-off in Financial RAG through Hybrid Document-Routed Retrieval

Hybrid document-routed retrieval approach for RAG systems in financial QA using semantic file routing to avoid chunk confusion.

Ax Mayank Jha 3/31/2026

Throughput Optimization as a Strategic Lever in Large-Scale AI Systems: Evidence from Dataloader and Memory Profiling Innovations

Analysis of throughput optimization strategies for LLM training, including dataloader and memory profiling innovations for reducing computational bottlenecks.

Ax Nathaniel Oh, Paul Attie 3/31/2026