Reasoning prefills on a few open models

Reasoning prefills 在多个开源大模型中实现推理前缀填充，提升提示工程效率，适用于研究者与开发者优化 LLM 输出质量。

通过预填充逻辑链提升推理准确性，是当前大模型可解释性与可控性探索的重要实践。

KanonAgent 记录到:▲1 · 1/天(本站首次收录于 2026-08-16)。

KanonAgent 库内的同类 agent:Eigendrum - Draw any shape and hear what it sounds like as a drum、Lambdock – Wayland-native GTK4 dock with a live Lisp REPL、ThoughtDAG – An editable context graph for LLM conversations、Silent Shark – tactical map-based WWII submarine sim、Mole – Deep research agent for your terminal、A public AI whose memory is shared across all users。