Wen Sun

All Roads Lead to Likelihood: The Value of Reinforcement Learning in Fine-Tuning

Prompt Curriculum Learning for Efficient LLM Post-Training

Expressive Value Learning for Scalable Offline Reinforcement Learning

Value-as-Return: A Two-Stage Framework to Align on the Optimal Score Function

Controllable Diffusion via Optimal Classifier Guidance

Test Time Scaling of Diffusion Model via Flow Matching Corrector

Computationally Efficient RL under Linear Bellman Completeness for Deterministic Dynamics

$Q\sharp$: Provably Optimal Distributional RL for LLM Post-Training

Accelerating RL for LLM Reasoning with Optimal Advantage Regression

$Q\sharp$: Provably Optimal Distributional RL for LLM Post-Training

Value-Guided Search for Efficient Chain-of-Thought Reasoning

Model-based RL as a Minimalist Approach to Horizon-Free and Second-Order Bounds

On Speeding Up Language Model Evaluation

Avoiding exp(R) scaling in RLHF through Preference-based Exploration

Diffusing States and Matching Scores: A New Framework for Imitation Learning

Scaling Offline RL via Efficient and Expressive Shortcut Models

Efficient Imitation under Misspecification

Regressing the Relative Future: Efficient Policy Optimization for Multi-turn RLHF

Correcting the Mythos of KL-Regularization: Direct Alignment without Overoptimization via Chi-Squared Preference Optimization

A Reductions Approach to Risk-Sensitive Reinforcement Learning with Optimized Certainty Equivalents

Convergence Of Consistency Model With Multistep Sampling Under General Data Assumptions

On Orchestrating Personalized LLMs

Convergence of Consistency Model with Multistep Sampling under General Data Assumptions