3 ms·
There was a paper[1] from last year where the authors discovered getting the model to output anything during times of uncertainty, improved the generations over
by gardnr 2y ago
There was a paper[1] from last year where the authors discovered getting the model to output anything during times of uncertainty, improved the generations overall. If all of the post-training alignment reasoning starts with the same tokens then I could see how it would condition the model to continue the reasoning phase.
1: https://arxiv.org/abs/2404.15758 https://arxiv.org/abs/2404.15758
- throwawaymaths 2y agothis is probably because the thinking tokens have the opportunity to store higher level/summarized contextual reasoning (lookup table based associations) in those token's KV caches. so an "Ok so" in position X may contain summarization vibes that are distinct from that in position Y.