Retrieval-of-Thought: Efficient Reasoning via Reusing Thoughts

1. 1Organize prior reasoning steps into a thought graph with sequential and semantic edges
2. 2Enable fast retrieval of query-relevant nodes from the thought graph
3. 3Recombine retrieved nodes to guide new problem-solving
4. 4Integrate RoT into large reasoning models to improve inference-time efficiency

AI researchers and engineers on a team can benefit from RoT as it enables fast retrieval and flexible recombination of prior reasoning steps, improving model efficiency and reducing latency

💡 Reusing prior reasoning steps can significantly improve inference-time efficiency in large reasoning models

💡 Improve inference-time efficiency with Retrieval-of-Thought (RoT) by reusing prior reasoning steps!