zlacker

[return to "LLMs cannot find reasoning errors, but can correct them"]
1. agentu+P8[view] [source] 2023-11-20 20:08:25
>>koie+(OP)
This might deserve some context here from experts. Wouldn’t solving mistake finding, in the general case, be the same as solving SAT (NP-Hard)?

From the abstract it sounds to me like they’re talking about heuristics for particular problems. Is that accurate?

◧◩
2. helen_+ok[view] [source] 2023-11-20 20:53:12
>>agentu+P8
Computational complexity isn’t really related here. Complexity has to do with formal languages and asymptotics, this is about natural language and fixed size data sets.
[go to top]