logoalt Hacker News

torginustoday at 6:59 PM1 replyview on HN

Wouldn't just putting tokens in a ring buffer work?


Replies

amlutotoday at 7:37 PM

Not unless you want to cheat the attention mechanism or do extra computations running prefill in a front-truncated version of the conversation.

Also, to the extent that the model reasons and thus learns something, if you blindly truncate the front, you will lose that knowledge. In the OP, the LLM that actually navigated the course successfully only did it on the second try. It it forgot the first failed try, it might not have succeeded :)