iCLP: LLM Reasoning with Implicit Cognition Latent Planning
Published:Dec 30, 2025 06:19
•1 min read
•ArXiv
Analysis
This paper introduces iCLP, a novel framework to improve Large Language Model (LLM) reasoning by leveraging implicit cognition. It addresses the challenges of generating explicit textual plans by using latent plans, which are compact encodings of effective reasoning instructions. The approach involves distilling plans, learning discrete representations, and fine-tuning LLMs. The key contribution is the ability to plan in latent space while reasoning in language space, leading to improved accuracy, efficiency, and cross-domain generalization while maintaining interpretability.
Key Takeaways
Reference
“The approach yields significant improvements in both accuracy and efficiency and, crucially, demonstrates strong cross-domain generalization while preserving the interpretability of chain-of-thought reasoning.”