3 ms·
That’s more like an inherent limitation of autoregressive prediction than that of LLM. Maybe LLMs can be trained or finetuned in other ways that allows it to th
by renonce 3y ago
That’s more like an inherent limitation of autoregressive prediction than that of LLM. Maybe LLMs can be trained or finetuned in other ways that allows it to think before answering.
- jeremyjh 3y agoThinking before answer would just be like writing its text output into a hidden buffer that only it is including in the context of the conversation. So its like its always also having a private conversation with its self. I can't prove this isn't more or less what I am.