3 ms·
Yes, there is an LLM feature that we have anthropomorphized as "reasoning" or "thinking", where an LLM has a scratch space where it can dump tokens that help to
by arcanemachiner 3mo ago
Yes, there is an LLM feature that we have anthropomorphized as "reasoning" or "thinking", where an LLM has a scratch space where it can dump tokens that help to improve the final output.
- otabdeveloper4 3mo ago> that help to improve the final output Do they actually help? Are you sure?