Remix.run Logo
Codex reasoning-token clustering at 516 may be leading to degraded performance(github.com)
10 points by 0x_rs 6 hours ago | 1 comments
cyanydeez an hour ago | parent [-]

wonder if theyre basically doing this llamacpp reasoning trick: https://github.com/ggml-org/llama.cpp/blob/master/common/rea...

id guess its the harness