[2404.15758] Let's Think Dot by Dot: Hidden Computation in Transformer Language Models

Abstract:Chain-of-thought responses from language models improve performance across most benchmarks. However, it remains unclear to what extent these performance gains can be attributed to human-like task decomposition or simply the greater computation that additional tokens allow. We show that transformers can use meaningless filler tokens (e.g., '......') in place of a chain of thought to solve two hard algorithmic tasks they could not solve when responding without intermediate tokens. However, we find empirically that learning to use filler tokens is difficult and requires specific, dense supervision to converge. We also provide a theoretical characterization of the class of problems where filler tokens are useful in terms of the quantifier depth of a first-order formula. For problems satisfying this characterization, chain-of-thought tokens need not provide information about the intermediate computational steps involved in multi-token computations. In summary, our results show that additional tokens can provide computational benefits independent of token choice. The fact that intermediate tokens can act as filler tokens raises concerns about large language models engaging in unauditable, hidden computations that are increasingly detached from the observed chain-of-thought tokens.

Let’s Think Dot by Dot: Hidden Computation in Transformer Language Models Jacob Pfau, William Merrill & Samuel R. Bowman Center for Data Science New York University NY, NY 10012, USA {jp6263,willm,bowman}@nyu.edu arXiv:2404.15758v1 [cs.CL] 24 Apr 2024…

saved by

• Evie Hu

related reading

• [2510.09714] All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Languagearxiv.org
• 2404.15758arxiv.org
• Let's Think Dot by Dot: Hidden Computation in Transformer Language Modelsarxiv.org
• Training Large Language Models to Reason in a Continuous Latent Spacearxiv.org
• [2005.14165] Language Models are Few-Shot Learnersarxiv.org
• A Mathematical Framework for Transformer Circuitstransformer-circuits.pub
• Circuit Tracing: Revealing Computational Graphs in Language Modelstransformer-circuits.pub
• 2402.12875arxiv.org
• Can LLMs Be Computers?percepta.ai
• [2201.11903] Chain of Thought Prompting Elicits Reasoning in Large Language Modelsarxiv.org
• [2511.08579] Training Language Models to Explain Their Own Computationsarxiv.org
• Exact Expressive Power of Transformers with Paddingarxiv.org