vicgalle/distilled-self-critique

distilled Self-Critique refines the outputs of a LLM with only synthetic data

21
/ 100
Experimental

This helps AI developers and researchers refine the outputs of large language models (LLMs) to better align with specific safety, sentiment, or privacy goals. It takes an existing LLM and synthetic training data, then produces a fine-tuned LLM that generates responses more reliably following desired criteria, like avoiding harmful content or maintaining a positive tone. This is for those building or deploying LLM-powered applications.

No commits in the last 6 months.

Use this if you need to improve the reliability and control over an LLM's output for specific behaviors (e.g., safety, sentiment) without requiring extensive human-annotated feedback data.

Not ideal if you are looking for a general-purpose LLM alignment solution that doesn't involve fine-tuning or if your primary concern isn't related to output refinement via synthetic data.

LLM-alignment AI-safety sentiment-control privacy-enhancement model-fine-tuning
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 5 / 25
Maturity 16 / 25
Community 0 / 25

How are scores calculated?

Stars

11

Forks

Language

Jupyter Notebook

License

MIT

Last pushed

Apr 11, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/vicgalle/distilled-self-critique"

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