Reactions from ranked influencers
2 posts
Prompting “……………….” makes AI think more The tokens don’t matter, just giving it extra time to think and have compute loop over the other tokens is enough to be smarter Btw this is a big reason planning mode works. If you flood the context window, you get more compute by default
If you ask a frontier LLM a multi-hop reasoning question, e.g., "Who won the Nobel Prize for Chemistry in (1900 + Mozart's age when he died)?", it usually can't answer correctly immediately (no thinking) BUT if you ask the same question & append 300 dots, suddenly it can answer?
Now apply this logic to "CoT faithfulness" and see what happens (disclaim illusions)
If you ask a frontier LLM a multi-hop reasoning question, e.g., "Who won the Nobel Prize for Chemistry in (1900 + Mozart's age when he died)?", it usually can't answer correctly immediately (no thinking) BUT if you ask the same question & append 300 dots, suddenly it can answer?