A Note on Hybrid Online Reinforcement and Imitation Learning for LLMs: Formulations and Algorithms
This work addresses fine-tuning challenges for LLM developers, but it appears incremental as it builds on existing methods.
The paper tackles the problem of fine-tuning Large Language Models by integrating Imitation Learning and Reinforcement Learning into a unified framework, resulting in a decomposition into dense and sparse gradients for efficient GPU implementation.
We present a unified framework for Large Language Model (LLM) fine-tuning that integrates Imitation Learning and Reinforcement Learning. By analyzing the gradient of a composite objective combining trajectory-level KL divergence with task rewards, we derive a natural decomposition into two components: (1) an analytically computable Dense Gradient for token-level imitation, and (2) a Monte Carlo estimated Sparse Gradient for long-horizon reward optimization. The Dense Gradient admits a closed-form logit-level formula, enabling efficient GPU implementation.