Reproduce “Chain of Mindset: Reasoning with Adaptive Cognitive Modes” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tianyi Jiang, Arctanx An, Hengyi Feng, Naixin Zhai, Haodong Li, Xiaomin Yu, Jiahui Liu, Hanwen Du and 7 more
The Chain of Mindset (CoM) framework enables large language models to dynamically switch between distinct cognitive modes (mindsets) during problem-solving, mimicking human cognitive flexibility. It achieves state-of-the-art accuracy across diverse reasoning benchmarks, including mathematics and multimodal tasks, with efficient information management through a Context Gate.
Key insights
- Dynamic, step-level orchestration of diverse cognitive modes significantly enhances LLM reasoning capabilities, enabling human-like adaptability to complex problem demands.
- Specialized mindsets with isolated contexts are crucial for effectively addressing heterogeneous sub-tasks within a single problem-solving process.
- An adaptive Context Gate is essential for managing information flow between the meta-agent and specialized mindsets, improving both reasoning effectiveness and computational efficiency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.