Pending a better name, I think I would go with “Redwood’s ‘avoiding injurious completions’ project”.
I’d call it our language model adversarial training project, maybe? Your proposal seems fine too
Pending a better name, I think I would go with “Redwood’s ‘avoiding injurious completions’ project”.
I’d call it our language model adversarial training project, maybe? Your proposal seems fine too