Hacker News new | ask | show | jobs
Codex reasoning-token clustering at 516 may be leading to degraded performance (github.com)
13 points by 0x_rs 31 days ago
2 comments
cyanydeez 31 days ago
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

link
virengupta45 29 days ago
this is happening on xhigh config!! extremely pissing off
link