Ax Thai-Khanh Nguyen, Ngoc-Bich-Uyen Vo, Thieu N. Vo, Tan M. Nguyen, Cuong Pham 14d ago

MuonSSM: Orthogonalizing State Space Models for Sequence Modeling

Introduces MuonSSM framework stabilizing state space models for long-sequence modeling by conditioning update geometry rather than recurrent weights.

Ax Davide Domini, Gianluca Aguzzi, Ivana Dusparic, Danilo Pianini, Mirko Viroli 14d ago

Discovering Collaboration from Novelty: Random Network Distillation for Clustered Federated Learning

Proposes lightweight approach for clustered federated learning using random network distillation to discover client collaborations without coupling cluster assignment to training.

Ax Mark Rhee, Jamie Simon, Dhruva Karkada 14d ago

Muon learns balanced solutions in matrix factorization without slow saddle-to-saddle dynamics

Studies Muon optimizer dynamics on matrix factorization problems, showing it avoids slow saddle-to-saddle transitions compared to gradient descent.

Ax Srinivasa Rao P., Vangmayi P Reddy 14d ago

Informational Frustration in Neural Manifolds: Shannon Bottlenecks and the Limits of Learnability

Theoretical framework linking information theory and topology to explain generalization in overparameterized deep networks, addressing theory-practice gap.

Ax Woojoo Na, Jennifer Dy 14d ago

ITSPACE: Monotone Gaussian Optimal Transport Updates

Proposes ITSPACE algorithm for optimal transport updates on covariance matrices using Bures-Wasserstein distance for domain adaptation.

Ax Matan Schliserman, Gon Buzaglo, Itay Evron, Daniel Soudry 14d ago

Convergence of Continual Learning in Homogeneous Deep Networks

Theoretical analysis of convergence properties in continual learning with deep networks, characterizing sequential projections onto task margin sets.

Ax Kan Zhu, Mathew Jacob, Chenxi Ma, Yi Pan, Stephanie Wang, Arvind Krishnamurthy, Baris Kasikci 14d ago

TraceLab: Characterizing Coding Agent Workloads for LLM Serving

TraceLab characterizes real-world coding agent workloads and LLM serving patterns across multiple models for systems optimization.

Ax Ting-Wen Ko, Jonas Geiping 14d ago

Attractor States Emerge in Multi-Turn LLM Conversations

Study analyzing attractor state emergence in multi-turn LLM conversations, showing topic-independent stable behaviors in debate interactions.

Ax Mohit Raghavendra, Anisha Gunjal, Aakash Sabharwal, Yunzhong He 14d ago

SWE-INTERACT: Reimagining SWE Benchmarks as User-Driven Long-Horizon Coding Sessions

SWE-Interact testbed evaluating coding agents on multi-turn interactive tasks with progressive user requirements instead of complete upfront specifications.

Ax Haoran Jin, Xiting Wang, Shijie Ren, Hong Xie, Defu Lian 14d ago

C$^{2}$R: Cross-sample Consistency Regularization Mitigates Feature Splitting and Absorption in Sparse Autoencoders

Cross-sample Consistency Regularization method addressing feature splitting and absorption problems in Sparse Autoencoders for LLM interpretation.

Ax Subramanyam Sahoo, Aman Chadha, Vinija Jain, Divya Chaudhary 14d ago

Pessimism's Paradox: Conservative Offline Training Amplifies Reward Hacking During Online Adaptation in Reasoning Models

Study showing conservative offline training paradoxically amplifies reward hacking in reasoning models during online adaptation with DPO.

Ax Philip Zmushko, Egor Petrov, Nursultan Abdullaev, Mikhail Khrushchev, Samuel Horv\'ath 14d ago

One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining

Analysis showing one-step gradient delay doesn't hinder asynchronous pipeline parallelism for large-scale LLM pretraining with PipeDream-2BW.

Ax Xuanfan Ni, Liyan Xu, Chenyang Lyu, Longyue Wang, Mo Yu, Lemao Liu, Fandong Meng, Jie Zhou, Piji Li 14d ago

ReFreeKV: Towards Threshold-Free KV Cache Compression

ReFreeKV method for KV cache compression in LLM inference without requiring pre-determined domain-specific thresholds.

Ax Daoming Wan, Yizheng Huang, Jimmy X. Huang 14d ago

TextClusterLab: An Integrated Framework for Reliable Text Clustering Studies

TextClusterLab framework for reliable evaluation of text clustering algorithms addressing dataset quality and semantic boundary challenges.

Ax Yichuan Wang, Zhifei Li, Zirui Wang, Paul Teiletche, Lesheng Jin, Matei Zaharia, Joseph E. Gonzalez, Sewon Min 14d ago

PIXELRAG: Web Screenshots Beat Text for Retrieval-Augmented Generation

PixelRAG method for retrieval-augmented generation using website screenshots in pixel space instead of parsed text for improved context.

Ax Charles L. Wang, Keir Dorchen, Peter Jin 14d ago

Agentic Safety is an Epistemic Property, Not a Behavioral One

Argues AI agent safety is epistemic property dependent on system correctability during learning, not just current behavior snapshots.

Ax Irene Strauss, Alexandra Butoi, Ryan Cotterell 14d ago

Generating in the Limit with Infinitely Many Hallucinations

Theoretical framework studying language generation in the limit and hallucinations as unavoidable consequence of learning.

Ax Yuyan Qian, Claude Montacie, Milan Stankovic, Victoria Eyharabide 14d ago

DBpedia-Enriched Company Representation for B2B Lead Recommendation

Approach enriching company embeddings with DBpedia semantic knowledge for B2B lead recommendation systems.

Ax Charith Chandra Sai Balne, Rithwik Maramraju, Siddharth Pratap Singh, Rohit Upadhyay, Aditya Singh, Chittaranjan Tripathy, Yogananda Domlur Seetharama 14d ago

The Voronoi Bottleneck: Capacity-Aware Dense Retrieval for Product Search

Theoretical analysis of capacity limits in dense retrieval using Voronoi geometry for product search with fixed embedding dimensions.

Ax Santosh Jaiswal 14d ago

Zero-Label Driving Scenario Complexity Detection via Joint Embedding Predictive Architecture

Unsupervised method for detecting complex driving scenarios using Joint Embedding Predictive Architecture without labels.

Ax Jiasheng Wang, Tanun Jitwatcharakomol, Piyawadee Jongpradubgiat, Simeng Zhu 14d ago

RADIANT-PET: Reasoning-Augmented PET/CT Lesion Segmentation with Large Language Models and Reinforcement Learning

RADIANT-PET framework combining segmentation models with LLM adjudication for improved lesion segmentation in PET/CT medical imaging.

Ax L. A. Mu\~noz 14d ago

GPU-Accelerated Inverse Structural Anastylosis from Block Collapse Dynamics

GPU-accelerated computational method for reassembling fallen stone blocks into original architectural configurations using inverse dynamics.

Ax Emily Bejerano, Federico Tondolo, Devang Gupta, Aaron Mano Cherian, Taeyoo Kim, Ayaan Qayyum, Xiaofan Yu, Xiaofan Jiang 14d ago

RadarTwin: Scene-Specific mmWave Radar Simulation and Learning for Mobile Indoor Perception

RadarTwin framework for generating synthetic mmWave radar training data using 3D reconstruction and vision-language models for mobile perception.

Ax Can Demircan, Marcel Binz, Alireza Modirshanechi, Eric Schulz 14d ago

Meta-learning as a principle for human-like visual representations

Research proposing meta-learning as principle for human-like visual representations in neural networks to support open-ended task flexibility.

Ax Yunhun Nam, Jongheon Jeong 14d ago

Vision-driven Preference Synthesis for Mitigating Hallucinations in VLMs

Method to reduce hallucinations in Vision-Language Models using preference alignment constructed from vision-driven synthesis rather than intervention-based approaches.

Ax Bruno Caro-V\'asquez, Carola Figueroa-Flores, Gast\'on Marquez 14d ago

Reinforcement Learning for Software Vulnerability Analysis: A Systematic Review with Emphasis on C/C++ Source Code and Static Analysis

Systematic review of reinforcement learning techniques for C/C++ vulnerability detection and static analysis following PRISMA guidelines.

Ax Xiangyu Qian, Jing Liu, Yunqing Tang, Luru Dai, Qiushi Li 14d ago

A Zero-Shot Deep Image Prior Framework for Denoising and Deconvolution in Fluorescence Microscopy

Zero-shot deep image prior framework for fluorescence microscopy denoising and deconvolution without paired training data.

Ax Rahid Zahid Alekberli, Hikmat Karimov 14d ago

Spectral Perturbation of the Empirical Fisher Information Matrix under Weight Quantization

Theoretical analysis of Fisher Information Matrix perturbations under quantization and distribution shift in parametric models.

Ax Jonghyeon Park, Olivier Jiyoun Jung, Myungwoo Oh 14d ago

LoRA-Tuned Large Language Models for Dementia Detection via Multi-View Speech-Derived Features

LoRA fine-tuning of LLMs for dementia detection using multi-modal speech features with automatic speech recognition transcripts.

Ax Yang Liu, Yuming Chen 14d ago

Event-Conditioned Diagnostics of Kinematic, Contact, and Object-Permanence Fields in Passive Object-State World Models

Research on how world models organize physical information in latent representations using diagnostic protocols for passive object-state prediction.

Ax Chanju Park, Dario Bocchi, Francesco D'Amico, Biagio Lucini, Gert Aarts 14d ago

Spectral phase transitions and trainability in neural network learning dynamics

Theoretical analysis of spectral phase transitions in neural network weight matrices during SGD training.

Ax Yuexi Du, Leya Barrientos, Laura Sheiman, John Lewin, Hemant D. Tagare, Nicha C. Dvornek 14d ago

MammoFlow: Multiview Mammogram Synthesis with Anatomically Consistent Flow Matching

Flow matching method for synthesizing paired mammogram views enforcing anatomical consistency.

Ax Kevin Der, Harish Kamath, Ben Thompson 14d ago

Turn-Averaged SAEs for Feature Discovery and Long-Context Attribution

Turn-averaged sparse autoencoders for interpretable feature extraction in language models with long context.

Ax Matteo Farina, Vishaal Udandarao, Thao Nguyen, Selim Kuzucu, Maximilian B\"other, Andreas Hochlehnert, Adhiraj Ghosh, Marianna Nezhurina, Karsten Roth, Joschka Struber, Yuhui Zhang, Sebastian Dziadzio, Elaine Sui, Soumya Jahagirdar, Dhruba Ghosh, Hasan Hammoud, Thomas De Min, Simone Caldarella, Jehanzeb Mirza, Sedrick Keh, Mehdi Cherti, Hilde Kuehne, Bernt Schiele, Serena Yeung-Levy, Muhammad Ferjad Naeem, Federico Tombari, Ana Klimovic, Elisa Ricci, Matthias Bethge, Sewoong Oh, Ameya Prabhu, Alessio Tonioni, Jenia Jitsev, Massimiliano Mancini, Ludwig Schmidt, Nikhil Parthasarathy 14d ago

DataComp-VLM: Improved Open Datasets for Vision-Language Models

DataComp benchmark for evaluating vision-language model dataset curation strategies with 160 open datasets.

Ax Chad A. Capps 14d ago

Depth-Staggered Fibonacci Spacing for Sparse Attention: Static Schedules Beat Learned Dilation and Extrapolate Where Dense Attention Fails

Study of sparse attention mechanisms using Fibonacci-spaced offsets in language models with depth-based scheduling.

Ax Arthur Josi, Emeline Got, Abdallah Dib, Luiz Gustavo Hafemann, Rafael M. O. Cruz 14d ago

KM-Speaker: Keypoint-Based Style Control for High-Quality Speech-Driven 3D Facial Animation and Dialogue Localization

Speech-driven 3D facial animation method with keypoint-based style control and dialogue localization.

Ax Aabesh Bhattacharyya, Tiffany Ding, Rina Foygel Barber 14d ago

Conformal Prediction with Macro-Coverage Guarantees

Conformal prediction method providing class-conditional and macro-level coverage guarantees for classification.

Ax Binh Nguyen, Colleen Josephson, Mircea Teodorescu, Gert Cauwenberghs, Jason Eshraghian 14d ago

Neuromorphic Energy-Aware Learning for Adaptive Deep Brain Stimulation

Energy-aware learning approach for neuromorphic computing in closed-loop deep brain stimulation systems.

Ax Oscar Mir\'o L\'opez-Feliu, Daimy van Loo, Xanthos Kekkos, Mikel Blom, Clara Rus 14d ago

Reproducing FACTER: Fairness via Conformal Thresholding and Prompt Repair

Reproducibility study of FACTER framework for fairness and coverage in LLM-based recommendation systems.

Ax Mohammad Imtiaz Hasan, Abyad Enan, Jean Michel Tine, Araf Rahman, M Sabbir Salek, Mashrur Chowdhury 14d ago

In-Vehicle Digital Twin-Based Collision Warning Framework with Sybil Attack Detection

Digital twin framework for connected vehicle collision warning with Sybil attack detection.

Ax Mudit Agarwal, Amit D. Bhrany 14d ago

Envisage: Diffusion-Based Rhinoplasty Goal Visualization with Mask-Decomposed Evaluation

Diffusion-based image editing pipeline for rhinoplasty visualization using FLUX.1 inpainting.

Ax Xin-She Yang, Mehmet Karamanoglu 14d ago

Analysis of Parameter Settings for the Bat Algorithm Using Variance Evolution

Theoretical analysis of parameter settings in bat algorithm metaheuristic using variance evolution.

Ax Zitian Zhou, Nan Lin 14d ago

Adaptive Iterative Hard Thresholding for Online High-dimensional Quantile Regression

Online sparse regression algorithm using adaptive iterative hard thresholding for high-dimensional quantile regression.

Ax Thiru Thillai Nadarasar Bahavan, Sachith Seneviratne, Saman Halgamuge 14d ago

FedLAS: Feature-Modulated Bidirectional Label Smoothing for Neural Network Calibration

Label smoothing technique for improving calibration of deep neural network classifiers.

Ax Aspen Erlandsson Brisebois, Luis Pablo Gonzalez Dominguez, Shivansi Prajapati, Zahed Khatooni, Heather L. Wilson, Connor Burbridge, Brook Byrns, Sureesh Tikoo, Christophe Pere, Steven Rayan, Gordon Broderick 14d ago

Exploring the Effects of Entanglement on Quantum Machine Learning of Pathogen Epitope-Receptor Binding

Study of quantum machine learning circuits for predicting epitope-receptor binding on NISQ devices.

Ax Aspen Erlandsson Brisebois, Zahed Khatooni, Connor Burbridge, Brook Byrns, Heather L. Wilson, Sureesh Tikoo, Steven Rayan, Gordon Broderick 14d ago

Transformer-Based Active Learning for Data-Efficient Vaccine Epitope Selection in PRRS

Transformer-based active learning approach for efficient vaccine epitope selection using molecular docking simulations.

Ax Andrea Carriero, Davide Pettenuzzo, Shubhranshu Shekhar 14d ago

MACROCAST: A Vintage-Consistent Time Series Foundation Model for Real-Time Macroeconomic Forecasting

Time series foundation model for macroeconomic forecasting addressing temporal contamination and revision bias issues.

Ax Eryn Libert-Scott, Emma A. M. Stanley, Vibujithan Vigneshwaran, Matthias Wilms, Erik Y. Ohara, Nils D. Forkert 14d ago

A Neuroimaging Simulation Framework for Developing and Evaluating Causal AI

Simulation framework for developing and testing causal AI methods on neuroimaging data for disease mechanism understanding.

Ax Dawon Ahn, Auder Der, Evangelos E. Papalexakis 14d ago

AnTenA: Actionable and Explainable Tensor Analysis System with Large Language Models

LLM-based system for explaining patterns in multi-aspect tensor data without requiring labels or metadata.

Ax Kazuki Ota, Takayuki Osa, Tatsuya Harada 14d ago

Self-Supervised Theorem Discovery in a Formal Axiomatic System

AI agent that autonomously discovers mathematical theorems in formal axiomatic systems without human priors, advancing machine reasoning capabilities.