Comment by formerly_proven
1 month ago
Correct. There is no token coloring. Models are just rl’d to attend to the first <systemprompt>…</systemprompt> strongly or “anything before token #4242”.
1 month ago
Correct. There is no token coloring. Models are just rl’d to attend to the first <systemprompt>…</systemprompt> strongly or “anything before token #4242”.
No comments yet
Contribute on Hacker News ↗