kv-cache-steering-reasoning

【效率与协作】Guide frozen language models toward multi-step reasoning by modifying cached key-value representations after the prefilling stage. Extract steering vectors from contrastive prompt pairs and apply them to KV cache with scalar coefficients. Improves reasoning on GSM8K, ARC, CommonsenseQA while adding only 10ms overhead per token.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准