Post #2890663

@atpfm@mastodon.social @siracusa@mastodon.social re:LLMs and deterministic outputs, your overall point stands, but I don't think JSON is a good example anymore - JSON is a great fit for constrained decoding, where you can deterministically force the model to only select tokens such that the output is not only valid JSON, but adherent to a specific schema. no try/retry needed! - see: https://platform.claude.com/docs/en/build-with-claude/structured-outputs https://developers.openai.com/api/docs/guides/structured-outputs