Poster
Self-Alignment with Instruction Backtranslation
Xian Li · Ping Yu · Chunting Zhou · Timo Schick · Omer Levy · Luke Zettlemoyer · Jason E Weston · Mike Lewis
Halle B #74
Fri 10 May 6:45 a.m. PDT — 7:30 a.m. PDT
We present a scalable method to build a high quality instruction following language model by automatically labelling human-written text with corresponding instructions. Our approach, named instruction backtranslation, starts with a language model finetuned on a small amount of seed data, and a given web corpus. The seed model is used to construct training examples by generating instruction prompts for web documents (self-augmentation), and then selecting high quality examples from among these candidates (self-curation). This data is then used to finetune a stronger model. Finetuning LLaMa on two iterations of our approach yields a model that outperforms all other LLaMa-based models on the Alpaca leaderboard not relying on distillation data, demonstrating highly effective self-alignment.