Ax Yuanhong Zheng, Ruichuan An, Xiaopeng Lin, Yuxing Liu, Sihan Yang, Huanyu Zhang, Haodong Li, Qintong Zhang, Renrui Zhang, Guopeng Li, Yifan Zhang, Yuheng Li, Wentao Zhang 3/24/2026

PEARL: Personalized Streaming Video Understanding Model

PEARL: streaming video understanding model for personalized real-time interactive responses from continuous visual input and instant feedback.

Ax Weili Cao, Xunjian Yin, Bhuwan Dhingra, Shuyan Zhou 3/24/2026

Coding Agents are Effective Long-Context Processors

Study showing coding agents effectively process long context by externalizing processing into explicit executable interactions rather than latent attention.

Ax Yen-Ting Piao, Jay Chiehen Liao, Wei-Tang Chien, Toshiki Ogimoto, Shang-Tse Chen, Yun-Nung Chen, Chun-Yi Lee, Shao-Yuan Lo 3/24/2026

ALICE: A Multifaceted Evaluation Framework of Large Audio-Language Models' In-Context Learning Ability

ALICE evaluation framework for assessing in-context learning ability of six large audio-language models under audio conditioning with reduced guidance.

Ax Truong Quynh Hoa, Hoang Dinh Cuong, Truong Xuan Khanh 3/24/2026

Detecting Neurovascular Instability from Multimodal Physiological Signals Using Wearable-Compatible Edge AI: A Responsible Computational Framework

Melaguard: lightweight Transformer edge AI framework (1.2M parameters) for detecting neurovascular instability from wearable physiological signals.

Ax Cailin Winston, Claris Winston, Ren\'e Just 3/24/2026

Solver-Aided Verification of Policy Compliance in Tool-Augmented LLM Agents

Solver-aided verification framework ensuring tool-augmented LLM agents comply with domain-specific operational policies for sensitive applications.

Ax Rounak Saha, Gurusha Juneja, Dayita Chaudhuri, Naveeja Sajeevan, Nihar B Shah, Danish Pruthi 3/24/2026

Policies Permitting LLM Use for Polishing Peer Reviews Are Currently Not Enforceable

Study showing LLM-detection policies for peer review are unenforceable; evaluates five detectors on human-AI collaboration scenarios.

Ax Yalemzerf Getnet, Abiy Tasissa, Waltenegus Dargie 3/24/2026

Shift-Invariant Feature Attribution in the Application of Wireless Electrocardiograms

Feature attribution method for ECG signal analysis to explain machine learning model decisions in biomedical applications.

Ax \c{S}uayp Talha Kocabay, Talha R\"uzgar Akku\c{s} 3/24/2026

Diffutron: A Masked Diffusion Language Model for Turkish Language

Diffutron: masked diffusion language model for Turkish using LoRA-based continual pretraining, addressing morphologically rich language modeling.

Ax Robert Godwin-Jones 3/24/2026

Profiling learners' affective engagement: Emotion AI, intercultural pragmatics, and language learning

Application of emotion AI and intercultural pragmatics to measure affective engagement in language learning contexts.

Ax Bernardo Magri, Benjamin Marsh, Paul Gebheim 3/24/2026

Meeting in the Middle: A Co-Design Paradigm for FHE and AI Inference

Co-design approach for fully homomorphic encryption and AI inference to enable privacy-preserving cloud model execution without weight updates.

Ax Dani Roytburg, Shreya Sridhar, Daphne Ippolito 3/24/2026

Measuring Reasoning Trace Legibility: Can Those Who Understand Teach?

Framework for evaluating legibility of reasoning traces in LLMs that output chain-of-thought deliberations alongside final answers.

Ax Anton Korikov, Scott Sanner 3/24/2026

ReBOL: Retrieval via Bayesian Optimization with Batched LLM Relevance Observations and Query Reformulation

ReBOL retrieval system using Bayesian optimization and LLM-based reranking to overcome vector similarity limitations and enable query reformulation.

Ax Mohammed Rakibul Hasan 3/24/2026

Evaluating Large Language Models on Historical Health Crisis Knowledge in Resource-Limited Settings: A Hybrid Multi-Metric Study

Evaluation of GPT-4, Gemini Pro, Llama-3, Mistral-7B on health crisis queries in low-resource Bangladesh context using multi-metric assessment.

Ax Ian Osband 3/24/2026

Delightful Distributed Policy Gradient

Distributed reinforcement learning approach addressing negative learning from high-surprisal data in stale or mismatched actor scenarios.

Ax Ian Osband 3/24/2026

Does This Gradient Spark Joy?

Policy gradient optimization method using 'delight' metric to selectively compute backward passes, reducing computation while maintaining learning value.

Ax Tony Mason 3/24/2026

Epistemic Observability in Language Models

Analysis of inverse correlation between model confidence and accuracy across Llama, Qwen, Mistral, and OLMo; formal proof that miscalibration is observational not capability gap.

Ax Ghislain Dorian Tchuente Mondjo 3/24/2026

Revenue-Sharing as Infrastructure: A Distributed Business Model for Generative AI Platforms

Business model analysis of generative AI platforms exploring multi-layer market architectures and revenue-sharing infrastructure for API/model providers.

Ax Muhammad Khalid, Yilmaz Uygun 3/24/2026

An Industrial-Scale Retrieval-Augmented Generation Framework for Requirements Engineering: Empirical Evaluation with Automotive Manufacturing Data

Industrial-scale RAG framework for requirements engineering in automotive manufacturing, with empirical evaluation on production data and performance metrics.

Ax Tianyi Huang, Nathan Huang, Justin Tang, Wenqian Chen, Elsa Fan 3/24/2026

Permutation-Consensus Listwise Judging for Robust Factuality Evaluation

PCFJudge: Inference-time method using permutation consensus to improve robustness of LLM-based factuality judges against candidate order sensitivity.

Ax Dong Liu, Yanxuan Yu, Ben Lengerich, Ying Nian Wu 3/24/2026

MKA: Memory-Keyed Attention for Efficient Long-Context Reasoning

MKA: Memory-Keyed Attention mechanism for efficient long-context LLM inference by reducing KV cache size while maintaining representation quality.

Ax Sen Fang, Weiyuan Ding, Zhezhen Cao, Zhou Yang, Bowen Xu 3/24/2026

AEGIS: From Clues to Verdicts -- Graph-Guided Deep Vulnerability Reasoning via Dialectics and Meta-Auditing

AEGIS: Graph-guided LLM agent for vulnerability detection using dialectics and meta-auditing to ground reasoning in code-specific evidence.

Ax Jon-Paul Cacioli 3/24/2026

Weber's Law in Transformer Magnitude Representations: Efficient Coding, Representational Geometry, and Psychophysical Laws in Language Models

Psychophysical analysis of how transformer language models represent magnitude using representational similarity, discrimination, and causal intervention.

Ax Ling Xiao, Toshihiko Yamasaki 3/24/2026

A Multihead Continual Learning Framework for Fine-Grained Fashion Image Retrieval with Contrastive Learning and Exponential Moving Average Distillation

Multihead continual learning framework for fine-grained fashion image retrieval using contrastive learning and exponential moving average distillation.

Ax Chien-Ping Lu 3/24/2026

Modernizing Amdahl's Law: How AI Scaling Laws Shape Computer Architecture

Study on how AI scaling laws and specialized accelerators reshape classical Amdahl's Law for modern heterogeneous computer architecture.

Ax Balaji Dinesh Gangireddi, Aniketh Garikaparthi, Manasi Patwardhan, Arman Cohan 3/24/2026

REVERE: Reflective Evolving Research Engineer for Scientific Workflows

REVERE: Reflective evolving research engineer agent that optimizes prompts and workflows using broader task patterns for scientific research coding.

Ax Tianyi Huang, Caden Yang, Emily Yin, Eric Wang, Michael Zhang 3/24/2026

PAVE: Premise-Aware Validation and Editing for Retrieval-Augmented LLMs

PAVE: Inference-time validation layer for retrieval-augmented LLMs that decomposes context into atomic facts to verify answer support.

Ax Jiarong Liang, Zhiheng Lyu, Zijie Liu, Xiangchao Chen, Ping Nie, Kai Zou, Wenhu Chen 3/24/2026

SWE-Next: Scalable Real-World Software Engineering Tasks for Agents

SWE-Next: Execution-grounded framework for scalable software engineering task collection to train AI agents on real-world code repositories.

Ax Fan Huang 3/24/2026

Reasoning Topology Matters: Network-of-Thought for Complex Reasoning Tasks

Network-of-Thought models LLM reasoning as directed graph instead of linear or tree structures for complex multi-source reasoning.

Ax Sunghyun Park, Jeongho Kim, Hyoungwoo Park, Debasmit Das, Sungrack Yun, Munawar Hayat, Jaegul Choo, Fatih Porikli, Seokeon Choi 3/24/2026

Memory-Efficient Fine-Tuning Diffusion Transformers via Dynamic Patch Sampling and Block Skipping

DiT-BlockSkip reduces memory during diffusion transformer fine-tuning through dynamic patch sampling and block skipping.

Ax Milos Gravara, Juan Luis Herrera, Stefan Nastic 3/24/2026

Compass: Optimizing Compound AI Workflows for Dynamic Adaptation

Compass optimizes compound AI workflows orchestrating multiple specialized models for accuracy, latency, and cost under dynamic loads.

Ax Wenjun Huang, Shenghao Fu, Yian Jin, Yang Ni, Ziteng Cui, Hanning Chen, Yirui He, Yezi Liu, Sanggeon Yun, SungHeon Jeong, Ryozo Masukawa, William Youngwoo Chung, Mohsen Imani 3/24/2026

MERIT: Multi-domain Efficient RAW Image Translation

MERIT addresses domain shifts in RAW images across camera sensors via efficient multi-domain translation model.

Ax Yujin Park, Haejun Chung, Ikbeom Jang 3/24/2026

Dodgersort: Uncertainty-Aware VLM-Guided Human-in-the-Loop Pairwise Ranking

Dodgersort uses CLIP-based ranking with uncertainty decomposition and information-theoretic sampling to reduce pairwise comparison labeling cost.

Ax Xiangyu Zeng, Qi Xu, Yunke Wang, Chang Xu 3/24/2026

HiCI: Hierarchical Construction-Integration for Long-Context Attention

HiCI proposes hierarchical attention module for long-context language modeling using segment-level representations and global integration.

Ax Shanshan Wang, Derek F. Wong, Jingming Yao, Lidia S. Chao 3/24/2026

Can ChatGPT Really Understand Modern Chinese Poetry?

Evaluates ChatGPT's comprehension of modern Chinese poetry through framework developed with professional poets.

Ax Saken Tukenov 3/24/2026

SozKZ: Training Efficient Small Language Models for Kazakh from Scratch

Develops SozKZ, family of 50M-600M parameter Llama models trained from scratch on Kazakh with dedicated tokenizer.

Ax Xander Coetzer, Arn\'e Schreuder, Anna Sergeevna Bosman 3/24/2026

Restoring Neural Network Plasticity for Faster Transfer Learning

Addresses neural network saturation during transfer learning by restoring plasticity of pretrained weights for downstream tasks.

Ax Hossein Javidnia 3/24/2026

Semantic Sections: An Atlas-Native Feature Ontology for Obstructed Representation Spaces

Introduces semantic sections as feature ontology for interpreting neural networks in obstructed representation spaces.

Ax Kaustubh D. Dhole, Eugene Agichtein 3/24/2026

RubricRAG: Towards Interpretable and Reliable LLM Evaluation via Domain Knowledge Retrieval for Rubric Generation

RubricRAG generates interpretable rubrics for LLM evaluation using domain knowledge retrieval to replace opaque scalar scoring.

Ax Zhaoyi Liu, Haichuan Zhang, Ang Li 3/24/2026

Beyond the Birkhoff Polytope: Spectral-Sphere-Constrained Hyper-Connections

Proposes manifold-constrained hyper-connections to stabilize training of models with multi-stream residual architectures.

Ax Joe Khawand, David Colliaux 3/24/2026

Natural Gradient Descent for Online Continual Learning

Applies natural gradient descent to online continual learning for image classification to prevent catastrophic forgetting.

Ax Hongyu Cao, Kunpeng Liu, Dongjie Wang, Yanjie Fu 3/24/2026

Mitigating Shortcut Reasoning in Language Models: A Gradient-Aware Training Approach

Proposes SART, a gradient-aware training method to detect and mitigate shortcut reasoning in LLMs through gradient surgery and scoring.

Ax Mohamed Aymen Bouyahia, Argyris Kalogeratos 3/24/2026

Enhancing LIME using Neural Decision Trees

Enhances LIME interpretability framework using neural decision trees for better explanation of complex ML models on tabular data.

Ax Lisan Al Amin, Md Ismail Hossain, Rupak Kumar Das, Mahbubul Islam, Saddam Mukta, Abdulaziz Tabbakh 3/24/2026

Democratizing AI: A Comparative Study in Deep Learning Efficiency and Future Trends in Computational Processing

Benchmarks deep learning models across GPUs to study computational efficiency and access disparities in training large-scale models.

Ax Arun Vignesh Malarkkan, Xinyuan Wang, Kunpeng Liu, Denghui Zhang, Yanjie Fu 3/24/2026

Causally-Guided Diffusion for Stable Feature Selection

CGDFS method for stable feature selection using causal principles to identify features robust to distribution shifts.

Ax Reshabh K Sharma, Dan Grossman 3/24/2026

AC4A: Access Control for Agents

AC4A access control framework enabling fine-grained permission management for LLM agents interacting with external APIs and tools.

Ax Yuren Hao, Shuhaib Mehri, ChengXiang Zhai, Dilek Hakkani-T\"ur 3/24/2026

User Preference Modeling for Conversational LLM Agents: Weak Rewards from Retrieval-Augmented Interaction

VARS framework for persistent user preference modeling in conversational LLM agents using retrieval-augmented interaction.

Ax Uchi Uchibeke 3/24/2026

Before the Tool Call: Deterministic Pre-Action Authorization for Autonomous AI Agents

Deterministic pre-action authorization framework for AI agent tool calls enforcing permission-based policy before execution.

Ax Jason Dury 3/24/2026

Beyond Expression Similarity: Contrastive Learning Recovers Functional Gene Associations from Protein Interaction Structure

Contrastive learning approach for discovering functional gene associations from protein interaction networks.

Ax Xinyue Liu, Niloofar Mireshghallah, Jane C. Ginsburg, Tuhin Chakrabarty 3/24/2026

Alignment Whack-a-Mole : Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models

Demonstrates finetuning bypasses safety alignment in LLMs, activating verbatim recall of copyrighted training data.

Ax Kemal Kirtac 3/24/2026

Learning to Aggregate Zero-Shot LLM Agents for Corporate Disclosure Classification

Multi-agent framework aggregating zero-shot LLM judgments for corporate disclosure classification using trained lightweight aggregator.