Ax Paul-Tiberiu Iordache, Elena Burceanu 3d ago

Fine-Tuning Regimes Define Distinct Continual Learning Problems

Analysis showing continual learning performance varies significantly based on fine-tuning regime (trainable parameter subspace).

Ax Yang Fu, Peng Qin, Liming Chen, Zihao Zhang, Hao Yu, Yifei Wang 3d ago

Joint Energy Management and Coordinated AIGC Workload Scheduling for Distributed Data Centers: A Diffusion-Aided Reward Shaping Approach

Reinforcement learning approach for scheduling AIGC workloads and managing energy in distributed data centers using diffusion-based reward shaping.

Ax Elahe Khatibi, Ziyu Wang, Saba A. Farahani, Di Huang, Hung Cao, Ramesh Jain, Amir M. Rahmani 3d ago

PerCaM-Health: Personalized Dynamic Causal Graphs for Healthcare Reasoning

PerCaM-Health method for learning personalized causal graphs from healthcare data with temporal dynamics and patient-level variation.

Ax Amir Mousavi, Mohammad Sadegh Sirjani, Erfan Nourbakhsh, Mimi Xie, Rocky Slavin, Leslie Neely, John Davis, John Quarles 3d ago

MambaGaze: Bidirectional Mamba with Explicit Missing Data Modeling for Cognitive Load Assessment from Eye-Gaze Tracking Data

MambaGaze framework using bidirectional Mamba architecture for cognitive load assessment from eye-tracking data with missing data handling.

Ax Noah Farr, Aryaman Reddi, Carlo D'Eramo, Jan Peters 3d ago

Streaming Reinforcement Learning under Partial Observability with Real-Time Recurrent Learning

Streaming reinforcement learning method enabling online learning with partial observability using real-time recurrent backpropagation.

Ax Zili Zhang, Chengxu Yang, Shenglong Zhang, Chenyu Wang, Yufan Zhang, Tuo Dai, Zhouyang Li, Yuhong Ge, Chao Jin, Xin Jin, Yuliang Liu 3d ago

BigMac: Breaking the Pareto Frontier of Compute and Memory in Multimodal LLM Training

BigMac training pipeline for multimodal LLMs that improves compute-memory efficiency tradeoffs through nested encoder-generator computation.

Ax Jason Sulskis, Sathya Ravi 3d ago

GENERIC-FNO: Embedding Energy Conservation and Entropy Production into Fourier Neural Operators

Fourier Neural Operator embedding full GENERIC thermodynamic structure including energy conservation and entropy production.

Ax Zhiwei Li, Yong Hu 3d ago

SkillHone: A Harness for Continual Agent Skill Evolution Through Persistent Decision History

Framework for continual evolution of agent skills in LLM-based agents by maintaining persistent decision history across task changes.

Ax Jos\'e A. Ch\'avez 3d ago

On the Redundancy of Timestep Embeddings in Diffusion Models

Analysis of redundancy in timestep embeddings within diffusion models, showing conditions where they may be unnecessary.

Ax Ananth K Suresh, Arya Hariharan 3d ago

Reproducibility Study of "AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models"

Reproducibility study of AlphaEdit null-space constrained knowledge editing method for LLMs, validating and extending original results.

Ax Jian Xu, Artur Miroszewski, John Paisley, Delu Zeng, Qibin Zhao 3d ago

Active Quantum Kernel Acquisition for Gaussian Process Regression

Active learning method for efficient quantum kernel estimation in Gaussian process regression with shot-budgeted quantum hardware.

Ax Ranuga Weerasekara, Heshan Nethmina, Manuja Ranathunga, Vinma Wettasinghe, Dinithi Navodya, Subavarshana Arumugam, Nirasha Munasinghe, Nisansa de Silva, Sandareka Wickramanayake 3d ago

When Prices Double in a Week: Forecasting of Agricultural Volatility in Import-Isolated Markets

ML forecasting framework for agricultural price volatility in import-isolated markets using supply-chain-aware features.

Ax Kathan Shah 3d ago

Token Geometry

Lightweight optimizer exploiting gradient geometry of embedding tables and LM-heads for improved training efficiency across finetuning and pretraining.

Ax Haotian Xie, Junlin Chen, Mingkai Zheng, Lishan Yang, Zhao Zhang 3d ago

PHOENIX: Resilient LLM Training with Hot-Swapping via Zero-Overhead Checkpoint

Fault-tolerant LLM training system using zero-overhead checkpointing and hot-swapping for resilience across hardware failures.

Ax Jen-Yen Chang, Takayuki Osa, Tatsuya Harada 3d ago

Learning the Supports for Categorical Critic in Reinforcement Learning

Investigation of Gaussian Histogram Loss for learning value function distributions in distributional reinforcement learning.

Ax Joy Bose 3d ago

Rank-Order N-of-M Codes for Sparse Distributed Memory: Disentangling Representation and Learning Effects in Noise Robustness Against Contemporary Neuromorphic Architectures

Evaluation of rank-order N-of-M encoding for sparse distributed memory as episodic memory for continual learning in LLMs.

Ax Zijun Xie, Yuyang You, Yongzhi Li, Enlei Gong, Zeyu Chen, Quan Chen, Yanhua Cheng, Peng Jiang, Yadong Mu 3d ago

ACPO: Adaptive Credit Policy Optimization via Fine-Grained Surrogate Entropy

Credit assignment optimization for RL-based LLM reasoning using fine-grained surrogate entropy for token-level rewards.

Ax Teng-Ruei Chen 3d ago

How Much of the Routing Gap Is Real? Decomposing the Router-to-Oracle Gap into Reproducible Specialist Advantage and Single-Draw Label Noise

Analysis decomposing router-to-oracle performance gap in LLM routing, identifying label noise versus specialist advantage contributions.

Ax Varvara Nazarenko, Timur Lidzhiev, Alexander Tarakanov 3d ago

PIEFS: Physics-Informed Eigenfunction Features with Learnable Scaling

Supervised representation learning framework using spectral methods with learnable scaling based on physics-informed inductive bias.

Ax Jaeyeon Kim, Jewon Lee, Bo-Kyeong Kim 3d ago

Quantize the Target, Quantize the Drafter: Efficient Inference with Qwen3.5-4B

Efficient inference system combining quantization and speculative decoding for Qwen3.5-4B LLM on resource-constrained hardware.

Ax Rahim Iqbal, Asfi Ahamed, Izzath Nisfer, Shazan Shaheed, Muhammadu Ilham, Nathali Athukorala, Madara Mendis, Nisansa de Silva, Sandareka Wickramanayake 3d ago

Environmental Drivers of Respiratory Disease: A District Level Analysis

Panel study analyzing environmental drivers of respiratory disease admissions across Sri Lankan districts.

Ax Kyohei Suzuki, Konstantinos Slavakis 3d ago

Non-Convex Sparse Reinforcement Learning via Non-Monotone Inclusions

Feature selection in reinforcement learning using non-convex regularization with projected minimax concave penalty.

Ax Ilya Burenko, Dmitry Vetrov 3d ago

Beyond Modality Fusion: Deep Ensembles for Multimodal Classification

Deep ensemble approach for multimodal classification without explicit modality fusion, addressing modality imbalance.

Ax Arash Esshaghi, Siavash Es'haghi, Gholamreza Shahabadi, Alireza Moradi 3d ago

SMART: A Machine Learning and Monte Carlo Framework for Rapid Analysis of Stochastic Transistor Aging and Process Variation in Digital Circuits

ML and Monte Carlo framework for analyzing transistor aging and process variation effects in digital circuits.

Ax Renzhen Wang, De cai, Kaiwen Xiao, Xixi Jia, Xiao Han, Deyu Meng 3d ago

Label Hierarchy Transition: Delving into Class Hierarchies to Enhance Deep Classifiers

Hierarchical classification approach exploiting category correlations through label hierarchy transitions in deep learning.

Ax Gregor Kobsik, Isaak Lim, Leif Kobbelt 3d ago

Partial Symmetry Detection for 3D Geometry using Contrastive Learning with Geodesic Point Cloud Patches

Contrastive learning method for detecting partial symmetry in 3D geometry using geodesic point cloud patches.

Ax Jianyi Zhang, Hao Frank Yang, Ang Li, Xin Guo, Pu Wang, Haiming Wang, Yiran Chen, Hai Li 3d ago

MLLM-LLaVA-FL: Multimodal Large Language Model Assisted Federated Learning

Federated learning framework using multimodal LLMs (LLaVA) to address data heterogeneity across distributed clients.

Ax Seemanta Bhattacharjee, MD. Muhtasim Fuad, A. K. M. Fakhrul Hossain 3d ago

Classification of Financial Data Using Quantum Support Vector Machine

Application of quantum support vector machines to financial data classification using Dhaka Stock Exchange dataset.

Ax Joachim Tomasi, Sandrine Anthoine, Hachem Kadri 3d ago

Benign Overfitting with Quantum Kernels

Theoretical study of benign overfitting in quantum kernel methods for machine learning on quantum computers.

Ax Gojko Perovic, Nuno Ferreira Duarte, Atabak Dehban, Gon\c{c}alo Teixeira, Egidio Falotico, Jos\'e Santos-Victor 3d ago

HERB: Human-augmented Efficient Reinforcement learning for Bin-packing

Human-augmented reinforcement learning approach for 3D bin-packing in logistics, combining RL with human feedback to reduce training time.

Ax Alireza Furutanpey, Carmen Walser, Philipp Raith, Pantelis A. Frangoudis, Schahram Dustdar 3d ago

Leveraging Neural Graph Compilers in Machine Learning Research for Edge-Cloud Systems

Evaluation of neural network graph compilers across heterogeneous hardware platforms, showing how vendor-specific optimizations affect performance comparisons.

Ax Alina Wernick, Kristof Meding 3d ago

Position: EU AI Act's Research Exemptions Can Break the Publication Norms of Major AI Conferences

Position paper on EU AI Act research exemptions and potential conflicts with academic publication norms at major conferences.

Ax Gautam Jajoo, Atharva Pandey, Pranjal A Chitale, Saksham Agarwal 3d ago

MASCA: LLM based-Multi Agents System for Credit Assessment

Multi-agent LLM system for credit assessment that mirrors real-world decision-making processes in financial evaluation.

Ax Quentin Uhl, Tommaso Pavan, Julianna Gerold, Kwok-Shing Chan, Yohan Jun, Shohei Fujita, Aneri Bhatt, Yixin Ma, Qiaochu Wang, Hong-Hsi Lee, Susie Y. Huang, Berkin Bilgic, Ileana Jelescu 3d ago

Reduced NEXI protocol for the quantification of human gray matter microstructure on the Connectome 2.0 scanner

Optimized MRI protocol for measuring brain microstructure using diffusion imaging with explainable AI framework.

Ax Constantin Venhoff, Iv\'an Arcuschin, Philip Torr, Arthur Conmy, Neel Nanda 3d ago

Base Models Know How to Reason, Thinking Models Learn When

Analysis of reasoning behaviors in thinking language models using Sparse Autoencoders and model diffing techniques.

Ax Hoang Ta, Jonathan Scarlett 3d ago

A Fast Binary Splitting Approach for Non-Adaptive Learning of Erd\H{o}s--R\'enyi Graphs

Graph learning algorithm for Erdős-Rényi graphs using group queries with fast binary splitting approach.

Ax Eliabelle Mauduit, Elo\"ise Berthier, Andrea Simonetto 3d ago

No-Regret Gaussian Process Optimization of Time-Varying Functions

Gaussian Process bandit optimization method for time-varying black-box function optimization with no-regret guarantees.

Ax Soroush Baghernezhad, Elaheh Mohammadreza, Vinicius Prado da Fonseca, Ting Zou, Xianta Jiang 3d ago

A Comparative Study of EMG- and IMU-based Gesture Recognition at the Wrist and Forearm

Comparative study of EMG and IMU sensors for hand gesture recognition at wrist and forearm locations.

Ax Scott Thornton 3d ago

SecureCode: A Production-Grade Multi-Turn Dataset for Training Security-Aware Code Generation Models

Production-grade dataset of 2,185 multi-turn examples for training secure code generation models covering OWASP Top 10 and ML security.

Ax Yu Wu, Minsik Jeon, Jen-Hao Rick Chang, Oncel Tuzel, Shubham Tulsiani 3d ago

RayRoPE: Projective Ray Positional Encoding for Multi-view Attention

Positional encoding scheme for multi-view transformers that maintains SE(3)-invariance and handles geometric relationships.

Ax Julia Werner, Julius Oexle, Oliver Bause, Maxime Le Floch, Franz Brinkmann, Hannah Tolle, Jochen Hampe, Oliver Bringmann 3d ago

Reliable Mislabel Detection for Video Capsule Endoscopy Data

Method for detecting mislabeled annotations in medical imaging datasets for video capsule endoscopy classification.

Ax Jean-Fran\c{c}ois Giovannelli 3d ago

A Gibbs posterior sampler for inverse problem based on prior diffusion model

Gibbs sampling algorithm for posterior inference in inverse problems with diffusion model priors.

Ax Jean-Fran\c{c}ois Giovannelli 3d ago

Estimation of instrument and noise parameters for inverse problem based on prior diffusion model

Bayesian inverse problem solving method using diffusion model priors for parameter estimation.

Ax Joyjit Roy, Samaresh Kumar Singh 3d ago

Agentic AI for Commercial Insurance Underwriting with Adversarial Self-Critique

LLM-based agentic AI system for insurance underwriting with self-critique mechanisms for high-stakes decision support.

Ax Haitao Lin, Hanyang Yu, Jingshun Huang, He Zhang, Yonggen Ling, Ping Tan, Xiangyang Xue, Yanwei Fu 3d ago

PoseVLA: Universal Pose Pretraining for Generalizable Vision-Language-Action Policies

Vision-Language-Action model for robotic control that combines VLM representations with 3D pose understanding for embodied AI tasks.

Ax Shree Harsha Bokkahalli Satish, Harm Lameris, Joakim Gustafson, \'Eva Sz\'ekely 3d ago

What Counts as Real? Speech Restoration and Voice Quality Conversion Pose New Challenges to Deepfake Detection

Study on deepfake detection challenges posed by voice quality conversion and speech restoration transformations.

Ax Samuel Ofosu Mensah, Camila Roa, Kerol Djoumessi, Philipp Berens 3d ago

Towards Interpretable Foundation Models for Retinal Fundus Images

Foundation model for medical imaging with interpretable-by-design architecture using BagNet backbone for retinal fundus analysis.

Ax Tianjiao Yu, Xinzhuo Li, Muntasir Wahed, Jerry Xiong, Yifan Shen, Ying Shen, Ismini Lourentzou 3d ago

DreamPartGen: Semantically Grounded Part-Level 3D Generation via Collaborative Latent Denoising

3D object generation method that decomposes objects into semantically meaningful parts using text-to-3D techniques.

Ax Conrad Borchers, Valdemar \v{S}v\'abensk\'y, Sandesh K. Kafle, Kevin K. Tang, Jan Vykopal 3d ago

Multimodal Analytics of Cybersecurity Crisis Preparation Exercises: What Predicts Success?

Study on instructional alignment in cybersecurity training simulations using multimodal data and Bloom's taxonomy coding.

Ax Zhenhang Shang, Yingzhe Yu, Kani Chen 3d ago

Fine-Tuning Integrity for Modern Neural Networks: Structured Drift Proofs via Norm, Rank, and Sparsity Certificates

Cryptographic method to verify fine-tuned neural network models haven't deviated from claimed update procedures without accessing parameters.