OK, I can partly explain the LLM chess weirdness now
SMRTR summary
Large language models (LLMs) demonstrate varying chess abilities, with gpt-3.5-turbo-instruct playing at an advanced amateur level while other models struggle. Experiments show that using specific prompting techniques, like regurgitation and examples, can significantly improve chess performance in newer LLMs, though still not matching gpt-3.5-turbo-instruct's skill level.
SMRTR provides this summary for quick context. The original article belongs to Hacker News.
Read the original article