Ax Junyi Wu, Dan Li 24d ago

Tensor-Efficient High-Dimensional Q-learning

Tensor-based Q-learning approach to handle high-dimensional reinforcement learning by exploiting problem structure without neural networks.

Ax Omri Ben-Dov, Luiz F. O. Chamon 24d ago

Adaptive Symmetrization of the KL Divergence

Adaptive symmetrization of KL divergence for learning probability distributions with normalizing flows and energy-based models.

Ax Zhixiong Zhao, Fangxin Liu, Junjie Wang, Chenyang Guan, Zongwu Wang, Li Jiang, Haibing Guan 24d ago

SpecQuant: Spectral Decomposition and Adaptive Truncation for Ultra-Low-Bit LLMs Quantization

SpecQuant framework for ultra-low-bit LLM quantization using spectral decomposition and adaptive truncation for efficient device deployment.

Ax Chin-Chia Michael Yeh, Uday Singh Saini, Xin Dai, Xiran Fan, Shubham Jain, Yujie Fan, Jiarui Sun, Junpeng Wang, Menghai Pan, Yingtong Dou, Yuzhong Chen, Vineeth Rakesh, Liang Wang, Yan Zheng, Mahashweta Das 24d ago

TREASURE: The Visa Payment Foundation Model for High-Volume Transaction Understanding

TREASURE foundation model for payment transaction understanding and analysis with applications to anomaly detection.

Ax Thomas Norrenbrock, Timo Kaiser, Sovan Biswas, Neslihan Kose, Ramesh Manuvinakurike, Bodo Rosenhahn 24d ago

CHiQPM: Calibrated Hierarchical Interpretable Image Classification

CHiQPM provides global and local interpretability for image classification in safety-critical domains with hierarchical explanations.

Ax Chihyeon Song, Jaewoo Lee, Jinkyoo Park 24d ago

Adaptive Replay Buffer for Offline-to-Online Reinforcement Learning

Adaptive Replay Buffer (ARB) dynamically prioritizes data sampling in offline-to-online reinforcement learning to balance stability and asymptotic performance.

Ax Carla Crivoi, Radu Tudor Ionescu 24d ago

Machine Unlearning in the Era of Quantum Machine Learning: An Empirical Study

First empirical study of machine unlearning in hybrid quantum-classical neural networks and variational quantum circuits.

Ax Pritthijit Nath, Sebastian Schemm, Henry Moss, Peter Haynes, Emily Shuckburgh, Mark J. Webb 24d ago

Replacing Tunable Parameters in Weather and Climate Models with State-Dependent Functions using Reinforcement Learning

Reinforcement learning framework to learn weather/climate model parametrization schemes as state-dependent functions online instead of using fixed coefficients.

Ax James O'Neill, Robert Clancy, Mariia Matskevichus, Fergal Reid 24d ago

Low-Rank Key Value Attention

Low-Rank Key-Value (LRKV) attention reduces transformer KV cache memory by exploiting redundancy across attention heads with low-rank residuals.

Ax Md Nabi Newaz Khan, Abdullah Arafat Miah, Yu Bi 24d ago

BadImplant: Injection-based Multi-Targeted Graph Backdoor Attack

BadImplant introduces multi-targeted backdoor attacks against graph neural networks with injection-based mechanisms.

Ax Annemarie Jutte, Uraz Odyurt 24d ago

Explainable AI to Improve Machine Learning Reliability for Industrial Cyber-Physical Systems

Explainable AI methods to improve ML reliability and prevent unexpected behavior in industrial cyber-physical systems.

Ax Powei Chang, Jinpeng Zhang, Bowen Chen, Chenyu Wang, Chenlu Guo, Yixing Zhang, Yukang Gao, JianXiang Xiang, Yue Gao, Chaoqun Sun, Yiyi Chen, Dongying Kong 24d ago

SPICE: Submodular Penalized Information-Conflict Selection for Efficient Large Language Model Training

SPICE uses submodular optimization and Fisher information to select training data for efficient LLM instruction tuning while addressing gradient conflicts.

Ax J Rosser, Robert Kirk, Edward Grefenstette, Jakob Foerster, Laura Ruis 24d ago

Infusion: Shaping Model Behavior by Editing Training Data via Influence Functions

Infusion framework uses influence functions to craft training data perturbations that induce targeted model behavior changes, evaluated on vision and language tasks.

Ax Alex Morehead, Miruna Cretu, Antonia Panescu, Rishabh Anand, Maurice Weiler, Tynan Perez, Samuel Blau, Steven Farrell, Wahid Bhimji, Anubhav Jain, Hrushikesh Sahasrabuddhe, Pietro Lio, Tommi Jaakkola, Rafael Gomez-Bombarelli, Rex Ying, N. Benjamin Erichson, Michael W. Mahoney 24d ago

Zatom-1: A Multimodal Flow Foundation Model for 3D Molecules and Materials

Open-source foundation model for 3D molecular and materials modeling with both generative and predictive capabilities.

Ax Dennis Thumm, Ying Chen 24d ago

Interventional Time Series Priors for Causal Foundation Models

Interventional time series data generator for training causal foundation models on time series, extending prior-data fitted networks to temporal domains.

Ax Nicolas Deutschmann, Constance Ferragu, Jonathan D. Ziegler, Shayan Aziznejad, Eli Bixby 24d ago

EvoFlows: Evolutionary Edit-Based Flow-Matching for Protein Engineering

EvoFlows: variable-length protein sequence model using flow matching for protein engineering with native support for insertions, deletions, and mutations.

Ax Philip Bechtle, Lucie Flek, Philipp Alexander Jung, Akbar Karimi, Timo Saala, Alexander Schmidt, Matthias Schott, Philipp Soldin, Christopher Wiebusch, Ulrich Willemsen 24d ago

Shapes are not enough: CONSERVAttack and its use for finding vulnerabilities and uncertainties in machine learning applications

CONSERVAttack method for identifying vulnerabilities and systematic uncertainties in ML models applied to high-energy physics data analysis.

Ax Paolo Toccaceli 24d ago

CRPS-Optimal Binning for Univariate Conformal Regression

Non-parametric conformal regression method using optimal binning with CRPS loss for conditional distribution estimation.

Ax Xianyong Xu, Yuanjun Zuo, Zhihong Huang, Yihan Qin, Haoxian Xu, Leilei Du, Haotian Wang 24d ago

MR-ImagenTime: Multi-Resolution Time Series Generation through Dual Image Representations

MR-CDM: multi-resolution conditional diffusion framework for variable-length time series forecasting with hierarchical decomposition and adaptive embeddings.

Ax Zhantao Chen, Dongyi He, Jin Fang, Xi Chen, Yishuo Liu, Xiaozhen Zhong, Xuejun Hu 24d ago

Toward Personalized Darts Training: A Data-Driven Framework Based on Skeleton-Based Biomechanical Analysis and Motion Modeling

Data-driven sports training framework using skeleton-based biomechanical analysis and motion modeling for personalized dart coaching.

Ax Chin-Chia Michael Yeh 24d ago

Matrix Profile for Time-Series Anomaly Detection: A Reproducible Open-Source Benchmark on TSB-AD

Open-source benchmark and reproducible implementation of Matrix Profile methods for univariate and multivariate time-series anomaly detection.

Ax Chenxu Yang, Chuanyu Qin, Qingyi Si, Minghui Chen, Naibin Gu, Dingyu Yao, Zheng Lin, Weiping Wang, Jiaqi Wang, Nan Duan 24d ago

Self-Distilled RLVR

On-policy self-distillation approach for LLM training combining dense teacher signals with sparse verifiable rewards from environment feedback.

Ax Maharshi Savdhariya 24d ago

NativeTernary: A Self-Delimiting Binary Encoding with Unary Run-Length Hierarchy Markers for Ternary Neural Network Weights, Structured Data, and General Computing Infrastructure

NativeTernary: binary encoding format for ternary neural network weights achieving 2 bits per weight, 1.31x compression over GGUF for BitNet models.

Ax Jonas De Schouwer, Haitz S\'aez de Oc\'ariz Borde, Xiaowen Dong 24d ago

k-Maximum Inner Product Attention for Graph Transformers and the Expressive Power of GraphGPS

Proposes k-maximum inner product attention mechanism for graph transformers to reduce computational complexity while maintaining expressive power.

Ax Bohao Li, Tao Zou, Junchen Ye, Yan Gong, Bowen Du 24d ago

A Clinical Point Cloud Paradigm for In-Hospital Mortality Prediction from Multi-Level Incomplete Multimodal EHRs

Deep learning approach for clinical risk prediction from incomplete multimodal EHR data using point cloud paradigm to handle irregular sampling and missing modalities.

Ax James Hu, Mahdi Ghelichi 24d ago

Noise Immunity in In-Context Tabular Learning: An Empirical Robustness Analysis of TabPFN's Attention Mechanisms

Empirical robustness analysis of TabPFN's attention mechanisms for tabular in-context learning, examining noise immunity across heterogeneous datasets.

Ax Tijana Zrnic, Emmanuel J. Cand\`es 24d ago

Active Statistical Inference

Active inference methodology for ML-assisted data collection, using models to identify which points merit labeling under budget constraints for efficient learning.

Ax Sawyer Robertson, Zhengchao Wan, Alexander Cloninger 24d ago

Resistance Distance and Linearized Optimal Transport on Graphs

Studies linearization of discrete transportation distance on graphs, connecting optimal transport to graph structure and providing nonasymptotic analysis.

Ax Daniel Adelman, Cagla Keceli, Alba V. Olivares-Nadal 24d ago

Thompson Sampling for Infinite-Horizon Discounted Decision Processes

Develops Thompson Sampling theory for discounted infinite-horizon MDPs with Borel state/action spaces and unknown parameters using canonical probability space framework.

Ax Achraf Azize, Marc Jourdan, Aymen Al Marjani, Debabrota Basu 24d ago

Differentially Private Best-Arm Identification

Studies best-arm identification with differential privacy guarantees in local and central models for privacy-sensitive applications like clinical trials and hyperparameter tuning.

Ax Tirthankar Mittra 24d ago

Exploring Natural Language-Based Strategies for Efficient Number Learning in Children through Reinforcement Learning

RL framework studying how children learn numbers using base-ten blocks, investigating numerical cognition through reinforcement learning and neural networks.

Ax Dimitri Meunier, Zhu Li, Tim Christensen, Arthur Gretton 24d ago

Nonparametric Instrumental Regression via Kernel Methods is Minimax Optimal

Provides convergence analysis and minimax optimality guarantees for kernel instrumental variable regression in both identified and non-identified settings.

Ax Siddharth Chandak 24d ago

Non-Expansive Mappings in Two-Time-Scale Stochastic Approximation: Finite-Time Analysis

Finite-time analysis of two-time-scale stochastic approximation algorithms with non-expansive mappings for optimization, reinforcement learning, and control applications.

Ax Penghui Yang, Cunxiao Du, Fengzhuo Zhang, Haonan Wang, Tianyu Pang, Chao Du, Bo An 24d ago

LongSpec: Long-Context Lossless Speculative Decoding with Efficient Drafting and Verification

LongSpec accelerates LLM inference on long contexts via lossless speculative decoding with efficient drafting and verification, targeting agent-based applications.

Ax Timo Gierlich, Andreas Baumbach, Akos F. Kungl, Kevin Max, Mihai A. Petrovici 24d ago

Spike-based alignment learning solves the weight transport problem

Spike-based alignment learning resolves weight transport problem in neural networks, enabling local computation compatible with biological networks and neuromorphic hardware.

Ax Wenlun Zhang, Yunshan Zhong, Weiqi Yan, Shengchuan Zhang, Shimpei Ando, Kentaro Yoshioka 24d ago

AHCQ-SAM: Toward Accurate and Hardware-Compatible Post-Training Segment Anything Model Quantization

AHCQ-SAM addresses post-training quantization challenges for Segment Anything Model to enable efficient deployment on resource-constrained edge devices.

Ax Andrea Montanari, Viet Vu 24d ago

Computational bottlenecks for denoising diffusions

Analyzes computational bottlenecks in denoising diffusion models, examining efficiency of drift learning and sampling procedures for probability distribution approximation.

Ax Aabid Karim, Abdul Karim, Bhoomika Lohana, Matt Keon, Jaswinder Singh, Abdul Sattar 24d ago

Lost in Cultural Translation: Do LLMs Struggle with Math Across Cultural Contexts?

Study of 14 LLMs showing mathematical reasoning accuracy drops 0.3-5.9% when math problems are culturally contextualized, revealing model limitations beyond pure logic.

Ax Tim Schneider, Cristiana de Farias, Roberto Calandra, Liming Chen, Jan Peters 24d ago

Apple: Toward General Active Perception via Reinforcement Learning

Active Perception Learner (Apple) applies reinforcement learning to enable general active perception in robotic systems with sparse, local sensory information.

Ax Yuhao Wu, Yushi Bai, Zhiqiang Hu, Roy Ka-Wei Lee, Juanzi Li 24d ago

LongWriter-Zero: Mastering Ultra-Long Text Generation via Reinforcement Learning

LongWriter-Zero uses reinforcement learning to improve ultra-long text generation in LLMs, overcoming length limitations and quality degradation without relying on synthetic training data.

Ax Zhentong Shao, Jingtao Qin, Nanpeng Yu 24d ago

Neural Two-Stage Stochastic Optimization for Solving Unit Commitment Problem

Neural stochastic optimization method for solving two-stage unit commitment problems using deep networks to approximate recourse costs under high-dimensional uncertainty.

Ax K. Giannoukou, X. Zhu, S. Marelli, B. Sudret 24d ago

MF-GLaM: A multifidelity stochastic emulator using generalized lambda models

MF-GLaM develops a multifidelity stochastic emulator using generalized lambda models for simulating conditional probability distributions in scientific computing.

Ax Nikolai Warner, Wenjin Zhang, Hamid Badiozamani, Irfan Essa, Apaar Sadhwani 24d ago

AugLift: Depth-Aware Input Reparameterization Improves Domain Generalization in 2D-to-3D Pose Lifting

AugLift improves 3D pose estimation from 2D keypoints using depth-aware input reparameterization and foundation models for better domain generalization.

Ax Wangsong Yin, Daliang Xu, Mengwei Xu, Gang Huang, Xuanzhe Liu 24d ago

ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference

ShadowNPU optimizes on-device LLM inference by addressing quantization sensitivity in attention operators, enabling efficient NPU execution for privacy-preserving deployment.

Ax Jackson Eshbaugh, Chetan Tiwari, Jorge Silveyra 24d ago

Synthetic Homes: A Multimodal Generative AI Pipeline for Residential Building Data Generation under Data Scarcity

Generative AI pipeline for synthetic building data creation addressing scarcity in residential energy modeling datasets.

Ax Junsong Li, Jie Zhou, Bihao Zhan, Yutao Yang, Qianjun Pan, Shilian Chen, Tianyu Huai, Xin Li, Qin Chen, Liang He 24d ago

LifeAlign: Lifelong Alignment for Large Language Models with Memory-Augmented Focalized Preference Optimization

Framework enabling LLMs to maintain alignment across sequential preference updates without catastrophic forgetting using memory-augmented optimization.

Ax Hind Atbir, Farah Cherfaoui, Guillaume Metzler, Emilie Morvant, Paul Viallard 24d ago

PAC-Bayesian Bounds on Constrained f-Entropic Risk Measures

PAC-Bayesian generalization bounds using constrained f-entropic risk measures for handling subgroup imbalances and distributional shifts.

Ax Zaid Khan, Archiki Prasad, Elias Stengel-Eskin, Jaemin Cho, Mohit Bansal 24d ago

One Life to Learn: Inferring Symbolic World Models for Stochastic Environments from Unguided Exploration

Approach for learning symbolic world models from single-episode exploration in stochastic environments without human guidance.

Ax Yongji Wu, Xueshen Liu, Haizhong Zheng, Juncheng Gu, Beidi Chen, Z. Morley Mao, Arvind Krishnamurthy, Ion Stoica 24d ago

RLBoost: Harvesting Preemptible Resources for Cost-Efficient Reinforcement Learning on LLMs

Method for cost-efficient reinforcement learning on LLMs using preemptible cloud resources, optimizing rollout and training stages separately.

Ax Zihan Zhao, Kaushik Pendiyala, Masood Mortazavi, Ning Yan 24d ago

PULSE: Privileged Knowledge Transfer from Rich to Deployable Sensors for Embodied Multi-Sensory Learning

Framework for transferring knowledge from rich sensor modalities to deployable sensors in embodied AI systems using multi-sensory learning.