Off-Policy Actor-Critic with Sigmoid-Bounded Entropy for Real-World Robot Learning

Deploying reinforcement learning in the real world remains challenging due to sample inefficiency, sparse rewards, and noisy visual observations. Prior work leverages demonstrations and human feedback to improve learning efficiency and robustness. However, offline-to-online methods need large datasets and can be unstable, while VLA-assisted RL relies on large-scale pretraining and fine-tuning. As a result, a low-cost real-world RL method with minimal data requirements has yet to emerge. We introduce \textbf{SigEnt-SAC}, an off-policy actor-critic method that learns from scratch using a single expert trajectory. Our key design is a sigmoid-bounded entropy term that prevents negative-entropy-driven optimization toward out-of-distribution actions and reduces Q-function oscillations. We benchmark SigEnt-SAC on D4RL tasks against representative baselines. Experiments show that SigEnt-SAC substantially alleviates Q-function oscillations and reaches a 100\% success rate faster than prior methods. Finally, we validate SigEnt-SAC on four real-world robotic tasks across multiple embodiments, where agents learn from raw images and sparse rewards; results demonstrate that SigEnt-SAC can learn successful policies with only a small number of real-world interactions, suggesting a low-cost and practical pathway for real-world RL deployment.

From Reward Shaping to Q-Shaping: Achieving Unbiased Learning with LLM-Guided Knowledge

Q-shaping is an extension of Q-value initialization and serves as an alternative to reward shaping for incorporating domain knowledge to accelerate agent training, thereby improving sample efficiency by directly shaping Q-values. This approach is both general and robust across diverse tasks, allowing for immediate impact assessment while guaranteeing optimality. We evaluated Q-shaping across 20 different environments using a large language model (LLM) as the heuristic provider. The results demonstrate that Q-shaping significantly enhances sample efficiency, achieving a \textbf{16.87\%} improvement over the best baseline in each environment and a \textbf{253.80\%} improvement compared to LLM-based reward shaping methods. These findings establish Q-shaping as a superior and unbiased alternative to conventional reward shaping in reinforcement learning.

Enhancing Q-Learning with Large Language Model Heuristics

Q-learning excels in learning from feedback within sequential decision-making tasks but requires extensive sampling for significant improvements. Although reward shaping is a powerful technique for enhancing learning efficiency, it can introduce biases that affect agent performance. Furthermore, potential-based reward shaping is constrained as it does not allow for reward modifications based on actions or terminal states, potentially limiting its effectiveness in complex environments. Additionally, large language models (LLMs) can achieve zero-shot learning, but this is generally limited to simpler tasks. They also exhibit low inference speeds and occasionally produce hallucinations. To address these issues, we propose \textbf{LLM-guided Q-learning} that employs LLMs as heuristic to aid in learning the Q-function for reinforcement learning. It combines the advantages of both technologies without introducing performance bias. Our theoretical analysis demonstrates that the LLM heuristic provides action-level guidance. Additionally, our architecture has the capability to convert the impact of hallucinations into exploration costs. Moreover, the converged Q function corresponds to the MDP optimal Q function. Experiment results demonstrated that our algorithm enables agents to avoid ineffective exploration, enhances sampling efficiency, and is well-suited for complex control tasks.