Rectifying Reinforcement Learning for Reward Matching
The Generative Flow Network (GFlowNet) is a probabilistic framework in which an agent learns a stochastic policy and flow functions to sample objects with probability proportional to an unnormalized reward function. GFlowNets share a strong resemblance to reinforcement learning (RL), that typically aims to maximize reward, due to their sequential decision-making processes. Recent works have studied connections between GFlowNets and maximum entropy (MaxEnt) RL, which modifies the standard objective of RL agents by learning an entropy-regularized objective. However, a critical theoretical gap persists: despite the apparent similarities in their sequential decision-making nature, a direct link between GFlowNets and standard RL has yet to be discovered, while bridging this gap could further unlock the potential of both fields. In this paper, we establish a new connection between GFlowNets and policy evaluation for a uniform policy. Surprisingly, we find that the resulting value function for the uniform policy has a close relationship to the flows in GFlowNets. Leveraging these insights, we further propose a novel rectified policy evaluation (RPE) algorithm, which achieves the same reward-matching effect as GFlowNets, offering a new perspective. We compare RPE, MaxEnt RL, and GFlowNets in a number of benchmarks, and show that RPE achieves competitive results compared to previous approaches. This work sheds light on the previously unexplored connection between (non-MaxEnt) RL and GFlowNets, potentially opening new avenues for future research in both fields.
Code (0)
등록된 구현이 없습니다.
Tasks
Decision Makingreinforcement-learningReinforcement LearningReinforcement Learning (RL)Sequential Decision MakingSimilar Papers 제목 키워드 기반
Rectifying LLM Thought from Lens of Optimization
Recent advancements in large language models (LLMs) have been driven by their emergent reasoning capabilities, particularly through long chain-of-thought (CoT) prompting, which enables thorough exploration and deliberati…
Reinforcement LearningSelf-Generated Critiques Boost Reward Modeling for Language Models
Reward modeling is crucial for aligning large language models (LLMs) with human preferences, especially in reinforcement learning from human feedback (RLHF). However, current reward models mainly produce scalar scores an…
Matching with AffNet based rectifications
We consider the problem of two-view matching under significant viewpoint changes with view synthesis. We propose two novel methods, minimizing the view synthesis overhead. The first one, named DenseAffNet, uses dense aff…
Rectifying Shortcut Behaviors in Preference-based Reward Learning
In reinforcement learning from human feedback, preference-based reward models play a central role in aligning large language models to human-aligned behavior. However, recent studies show that these models are prone to r…
Reinforcement LearningRectifying homographies for stereo vision: analytical solution for minimal distortion
Stereo rectification is the determination of two image transformations (or homographies) that map corresponding points on the two images, projections of the same point in the 3D space, onto the same horizontal line in th…