LLM Reasoning and Pruning Methods (paper implementation)

Paper implementation of pruning techniques of reasoning thought chains, to delete redundant tokens which unnecessary hog up the KV cache memory but do not contribute much in the reasoning or logic building to get to the final solution.