GPT-4: Researchers from Carnegie Mellon University, NVIDIA, Ariel University, and Microsoft have developed SPRING, a Large Language Model (LLM)-based policy that outperforms Reinforcement Learning algorithms in complex game environments. By leveraging prior knowledge from academic papers and employing in-context chain-of-thought reasoning, SPRING surpasses state-of-the-art methods on the Crafter benchmark, achieving significant improvements in-game score and reward. The results highlight the potential of LLMs in complex tasks and suggest future advancements in visual-language models could address existing limitations.
Read more at MarkTechPost…