Rosmine announces Distribution Fine Tuning, a post-training technique that refines large language model output distributions to achieve 100% human-written scores on Pangram 3.3.2 tests

— Demo shows Pangram detector labeling output as fully human-written.

Original post

RO#1220@KRISHNANROHITOPRORosmine|@ROSMINE

I fixed why LLMs write so poorly, and I have a demo to prove it Announcing Distribution Fine Tuning (DFT): A post training step that fixes LLM writing Model outputs fooled pangram on 100% of test cases

9:07 AM · May 18, 2026 View on X

Reposted by

CT#570|@CHENHAOTAN