haozheji/exact-optimization

ICML 2024 - Official Repository for EXO: Towards Efficient Exact Optimization of Language Model Alignment

27
/ 100
Experimental

This project helps machine learning engineers improve how language models respond to user prompts, ensuring outputs better align with human preferences. It takes a pre-trained language model and either human-curated preference data or reward model scores, then produces a fine-tuned model that generates more desirable text. This is for developers or researchers who build and deploy large language models and want to refine their behavior.

No commits in the last 6 months.

Use this if you need to fine-tune a language model so its generated text is more aligned with specific human preferences, using an efficient optimization method.

Not ideal if you are looking for a no-code solution or a tool for general-purpose language model deployment without focusing on preference alignment.

language-model-fine-tuning AI-model-alignment natural-language-generation ML-developer-tools responsible-AI
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 8 / 25
Maturity 16 / 25
Community 3 / 25

How are scores calculated?

Stars

56

Forks

1

Language

Python

License

MIT

Last pushed

Jun 16, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/haozheji/exact-optimization"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.