Democratic Fine-Tuning (DFT)

Democratic Fine-Tuning (DFT) is an OpenAI-funded project of the Meaning Alignment Institute that uses values cards and a moral graph in a short democratic deliberation process to elicit people’s values and aggregate them into training data for fine-tuning language models, providing an alternative to Constitutional AI and standard RLHF.

Endorsements support Meaning Alignment Institute.

No endorsements yet.

Discussion

Start Date

-

End Date

-

Expected Duration

-

Funding Raised to Date

-