Looks like grokking could give better reasoning and generalization to LLMs, but I'm not sure how practical it would be to overfit a larger LLM See: https://arx

Looks like grokking could give better reasoning and generalization to LLMs, but I'm not sure how practical it would be to overfit a larger LLM See: https://arxiv.org/abs/2405.15071 https://arxiv.org/abs/2405.15071