I assume it’s largely a side effect from the final RL in post training?
replyThat’s the step that causes the most significant gains in agentic performance.
But the RL doesn’t care about anything except maximizing the score, so if you only score based on coding benchmarks, anything can happen to the writing style (as long as it doesn’t hurt the coding performance).
That’s why it often gets worse on models that simply had more RL post training from the same base.
Does Xiaomis approach help with this? They do all the post training steps at the same time instead of one by one, switch topics after a couple prompts so writing style is mixed with coding and tool use.
replyApparently it helps generalize skills between areas, which makes sense when you compare it to how humans learn but I don't know if it's the same for LLMs.
Reinforcement learning for specific use-cases like coding that degrade it's writing style... makes sense. Maybe it stands to reason later version of Opus were improved more by this sort of fine-tuning. Feels consistent with the observation of diminishing returns and worsening writing style. Wonder what changed (supposedly) in 5.5.
replyIt truly was bizarre. I've used every major model since 2022, and not a single one had a writing style as bad as Opus 5
replyFable 5 was pretty bad too, but they fixed it with 5.1. Now with Opus 5.5 it seems they fixed it as well
reply