3 ms·
I don’t think you have an accurate understanding of how LLMs work. https://arxiv.org/abs/2501.19393 https://arxiv.org/abs/2501.19393 These tokens DO extend th
by computerex 2y ago
I don’t think you have an accurate understanding of how LLMs work.
https://arxiv.org/abs/2501.19393 https://arxiv.org/abs/2501.19393
These tokens DO extend the thinking time. We are talking about causal autoregressive language models, and so these tokens can be used to guide the generation.