Benchmarked: MindControl for Llama.cpp

I recently shared the original MindControl PoC (and on github) - sampler-level guided reasoning budgets for llama.cpp, nudging the model with self-aware statements about its own thinking budget instead of just hard-truncating it. We received some great feedback, and the most common ask (fair e

中文全文翻译正在补齐，可先查看下方英文原文。

I recently shared the original MindControl PoC (and on github) - sampler-level guided reasoning budgets for llama.cpp, nudging the model with self-aware statements about its own thinking budget instead of just hard-truncating it. We received some great feedback, and the most common ask (fair e