2 ms·
Didn’t anthropic show that the models engage in a form of planning such that it is predicting a possible future subsequent tokens that then affects prediction o
by Merik 1y ago
Didn’t anthropic show that the models engage in a form of planning such that it is predicting a possible future subsequent tokens that then affects prediction of the next token:
https://transformer-circuits.pub/2025/attribution-graphs/biology.html#dives-poems https://transformer-circuits.pub/2025/attribution-graphs/bio...
- ACCount37 1y agoSure, an LLM can start "preparing" for token N+4 at token N. But that doesn't change that the token N can't "see" N+1. Causality is enforced in LLMs - past tokens can affect future tokens, but not the other way around.