Fetching the paper…

Reading the bibliography…

2024

Do LLMs Really Think Step-by-step In Implicit Reasoning?

Yu, Yijiong

Original
PDF
Post on X

Understand

It has been well-known that Chain-of-Thought can remarkably enhance LLMs' performance on complex tasks.

However, because it also introduces slower inference speeds and higher computational costs, many researches have attempted to use implicit CoT, which does not need LLMs to explicitly generate the intermediate steps.
However, the invisible reasoning process leaves us a doubt that, can implicit CoT really be equal to explicit CoT? Therefore, in this study, we address this question through experiments.
We probe the information of intermediate steps from the model's hidden states when it is either trained or prompted to perform implicit CoT.

Do LLMs Really Think Step-by-step In Implicit Reasoning? · Around

Built on

Thinking, fast and slow
Daniel Kahneman. 2011 · 2011
Earlier work this paper cites.
Chain-of-thought prompting elicits reasoning in large language models
Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022 · 2022
Earlier work this paper cites.
Implicit chain of thought reasoning via knowledge distillation
Original
Yuntian Deng, Kiran Prasad, Roland Fernandez, Paul Smolensky, Vishrav Chaudhary, and Stuart Shieber. 2023 · 2023
Earlier work this paper cites.

Similar

Towards Better Chain-of-Thought Prompting Strategies: A Survey
Original
Zihan Yu, Liang He, Zhen Wu, Xinyu Dai, and Jiajun Chen. 2023 · 2023
Cited alongside, same era.
From explicit CoT to implicit CoT: Learning to internalize CoT step by step
Original
Yuntian Deng, Yejin Choi, and Stuart Shieber. 2024 · 2024
Cited alongside, same era.
O1 Replication Journey: A Strategic Progress Report – Part 1
Original
Yiwei Qin, Xuefeng Li, Haoyang Zou, Yixiu Liu, Shijie Xia, Zhen Huang, Yixin Ye, Weizhe Yuan, Hector Liu, Yuanzhi Li, and Pengfei Liu. 2024 · 2024
Cited alongside, same era.
Physics of language models: Part 3.2, knowledge manipulation
Original
Zeyuan Allen-Zhu and Yuanzhi Li
Cited in the paper.
Grokked transformers are implicit reasoners: A mechanistic journey to the edge of generalization
Original
Boshi Wang, Xiang Yue, Yu Su, and Huan Sun
Cited in the paper.
Do large language models latently perform multi-hop reasoning?
Original
Sohee Yang, Elena Gribovskaya, Nora Kassner, Mor Geva, and Sebastian Riedel
Cited in the paper.

Then

Qwen2.5: A party of foundation models!
Qwen Team. 2024 · 2024
Closest in time.
Physics of Language Models: Part 2.1, Grade-School Math and the Hidden Reasoning Process
Original
Tian Ye, Zicheng Xu, Yuanzhi Li, and Zeyuan Allen-Zhu. 2024 · 2024
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…