entropy-guided-regeneration

【效率与协作】Monitor Shannon entropy in LLM token distributions to detect sudden uncertainty spikes, then trigger adaptive prompt consolidation to realign conversation context and improve accuracy by 56.6% and reliability by 35.3%.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准