Tag
Introduces Constrained Flow Optimization (CFO), a framework for fine-tuning generative flow models to maximize rewards while satisfying constraints in molecular design, with theoretical guarantees and experimental validation.
Introduces Vector Policy Optimization (VPO), a new RL method that handles vector-valued rewards to improve test-time scaling for LLMs, outperforming conventional scalar reward approaches.
This paper introduces MARBLE, a gradient-space optimization framework for multi-reward reinforcement learning fine-tuning of diffusion models, which harmonizes policy gradients without manual weighting.
Self-Distillation Zero (SD-Zero) is a novel training method that converts sparse binary rewards into dense token-level supervision through dual-role training where a model acts as both generator and reviser, achieving 10%+ improvements on math and code reasoning benchmarks with higher sample efficiency than RL approaches.