Heng Huang

Catalog-Native LLM: Speaking Item-ID dialect with Less Entanglement for Recommendation

Parallel-R1: Towards Parallel Thinking via Reinforcement Learning

Achieve Performatively Optimal Policy for Performative Reinforcement Learning

New Hybrid Fine-Tuning Paradigm for LLMs: Algorithm Design and Convergence Analysis Framework

Zeroth-Order Methods for Stochastic Nonconvex Nonsmooth Composite Optimization

Riemannian Zeroth-Order Gradient Estimation with Structure-Preserving Metrics for Geodesically Incomplete Manifolds

An Ensemble Framework for Unbiased Language Model Watermarking

Learning to Reason via Mixture-of-Thought for Logical Reasoning

Trade-off in Estimating the Number of Byzantine Clients in Federated Learning

Toward Bit-Efficient Dataset Condensation: A General Framework

Mitigating Data Heterogeneity Effect in Client-Reshuffling-Based Federated Learning

Model Correlation Detection via Random Selection Probing

A Multi-Fidelity Mixture-of-Expert Framework Integrating PDE Solvers and Neural Operators for Computational Fluid Dynamics

Analyzing and Evaluating Unbiased Language Model Watermark

Provably Learning Representations under Generalized Dependency Structure

An Unlearning-Enhanced General Framework for Test-Time Adaptation

Explore Data Left Behind in Reinforcement Learning for Reasoning Language Models

CoTGuard: Using Chain-of-Thought Triggering for Copyright Protection in Multi-Agent LLM Systems

Defending against Model Extraction for GNNs with Model Reprogramming

Improving Text-to-Image Generation with Input-Side Inference-Time Scaling

Surprise-Modulated Meta-Advantages in Reinforcement Learning: Towards Language-Neutral Post-Training for Code LLMs

Pruning Without Fine-Tuning: Dynamic Pruning of Autoregressive Image Generation Models to Mixtures of Experts

Layer-Wise Feedback Signals: Dynamic Regulation for Continual Learning

Leave No TRACE: Black-box Detection of Copyrighted Dataset Usage in Large Language Models via Watermarking

Understanding Catastrophic Interference On the Identifibility of Latent Representations

Modality-Balancing Preference Optimization of Large Multimodal Models by Adversarial Negative Mining

Learning to Reason About Code Insecurity: Composite-Reinforcement Fine-Tuning for Cognitive Alignment

Graph Attention with Knowledge-Aware Domain Adaptation for Drug-Target Interaction Prediction

Understanding Weak-to-Strong Generalization: A Spectral Analysis

Advancing Drug-Target Interaction Prediction via Graph Transformers and Residual Protein Embeddings

DDI-Aware Domain Adaptation for Cross-Domain Drug Combination Representation Learning via Contrastive Embedding

Meta-Reinforcement Learning for Compiler Optimization: A Kernel-Embedded CompilerLLM with Verified Assumptions and Practical Guarantees

Cost-Aware Contrastive Routing for LLMs

On the Optimal Construction of Unbiased Gradient Estimators for Zeroth-Order Optimization

Robust Distortion-Free Watermark for Autoregressive Audio Generation Models

Bilevel ZOFO: Efficient LLM Fine-Tuning and Meta-Training

Revisiting Zeroth-Order Optimization: Minimum-Variance Two-Point Estimators and Directionally Aligned Perturbations

Identification of Intermittent Temporal Latent Process

Robust Reinforcement Learning in Finance: Modeling Market Impact with Elliptic Uncertainty Sets

Not All Prompts Are Made Equal: Prompt-based Pruning of Text-to-Image Diffusion Models

Revisiting Convergence: Shuffling Complexity Beyond Lipschitz Smoothness

OmnixR: Evaluating Omni-modality Language Models on Reasoning across Modalities

Revisiting Convergence: A Study on Shuffling-Type Gradient Methods

Provably Mitigating Corruption, Overoptimization, and Verbosity Simultaneously in Offline and Online RLHF/DPO Alignment

Bilevel ZOFO: Bridging Parameter-Efficient and Zeroth-Order Techniques for Efficient LLM Fine-Tuning and Meta-Training

A Watermark for Order-Agnostic Language Models

Towards Optimal Multi-draft Speculative Decoding

Cross-Modal Few-Shot Learning: a Generative Transfer Learning Framework

De-mark: Watermark Removal in Large Language Models

LLaVA-Critic: Learning to Evaluate Multimodal Models

Escaping Saddle Point Efficiently in Minimax and Bilevel Optimizations

Data Efficient Continual Learning of Large Language Model

Pseudo- vs. True-Randomness: Rethinking Distortion-Free Watermarks of Language Models under Watermark Key Collisions

From Pixels to Prose: A Large Dataset of Dense Image Captions

OPTune: Efficient Online Preference Tuning

Achieve Performatively Optimal Policy for Performative Reinforcement Learning

A Fast Federated Method for Minimax Problems with Sequential Convergence Guarantees

Hybrid Fine-Tuning of LLMs: Theoretical Insights on Generalized Smoothness and Convergence

Sharper Analysis of Data Echoing and New Communication-Efficient Algorithm for Data Parallelism

Provably Mitigating Corruption, Overoptimization, and Verbosity Simultaneously in Offline and Online RLHF/DPO Alignment

Efficient Machine Unlearning for Deep Generative Models by Mitigating Optimization Conflicts

RAG$^C$: Towards Copyright Protection for Knowledge Bases of Retrieval-augmented Language Models

Rectified Robust Policy Optimization for Robust Constrained Reinforcement Learning without Strong Duality