Chain of Thought Empowers Transformers to Solve Inherently Serial Problems · HackerPapers discussions

Chain of Thought Empowers Transformers to Solve Inherently Serial Problems — mentioned by pama

Fri, 03 Jan 2025 19:07:20 GMT

I recall early, incomplete speculation about transformers not solving Boolean circuit value problems; what did you think of this work? https://arxiv.org/abs/2402.12875v3

Paper and discussion history. A mention is not an endorsement.

Chain of Thought Empowers Transformers to Solve Inherently Serial Problems — mentioned by nopinsight

Tue, 17 Sep 2024 01:46:42 GMT

In the words of an author: "What is the performance limit when scaling LLM inference? Sky's the limit. We have mathematically proven that transformers can solve any problem, provided they are allowed to generate as many intermediate reasoning tokens as needed. Remarkably, constant depth is sufficient. http://arxiv.org/abs/2402.12875 (ICLR 2024)" https://x.com/denny_zhou/status/1835761801453306089

Paper and discussion history. A mention is not an endorsement.

Chain of Thought Empowers Transformers to Solve Inherently Serial Problems — HN submission

Tue, 17 Sep 2024 00:23:17 GMT

Chain of Thought empowers transformers to solve inherently serial problems

261 points; 184 discussion replies on HN. Paper and source history. A submission is not an endorsement.