The Expressive Power of Transformers with Chain of Thought
Abstract
Recent theoretical work has identified surprisingly simple reasoning problems, such as checking if two nodes in a graph are connected or simulating finite-state machines, that are provably unsolvable by standard transformers that answer immediately after reading their input. However, in practice, transformers' reasoning can be improved by allowing them to use a "chain of thought" or "scratchpad", i.e., generate and condition on a sequence of intermediate tokens before answering. Motivated by this, we ask: Does such intermediate generation fundamentally extend the computational power of a decoder-only transformer? We show that the answer is yes, but the amount of increase depends crucially on the amount of intermediate generation. For instance, we find that transformer decoders with a logarithmic number of decoding steps (w.r.t. the input length) push the limits of standard transformers only slightly, while a linear number of decoding steps, assuming projected pre-norm (a slight generalization of standard pre-norm), adds a clear new ability (under standard complexity conjectures): recognizing all regular languages. Our results also imply that linear steps keep transformer decoders within context-sensitive languages, and polynomial steps with generalized pre-norm make them recognize exactly the class of polynomial-time solvable problems -- the first exact characterization of a type of transformers in terms of standard complexity classes. Together, this provides a nuanced framework for understanding how the length of a transformer's chain of thought or scratchpad impacts its reasoning power.
- Publication:
-
arXiv e-prints
- Pub Date:
- October 2023
- DOI:
- 10.48550/arXiv.2310.07923
- arXiv:
- arXiv:2310.07923
- Bibcode:
- 2023arXiv231007923M
- Keywords:
-
- Computer Science - Machine Learning;
- Computer Science - Computational Complexity;
- Computer Science - Computation and Language;
- Computer Science - Logic in Computer Science
- E-Print:
- 9-page preprint. ICLR camera ready posted April 11