HippoSpark: An On-Demand Experience System for LLM Reasoning
2026-06-29 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors present HippoSpark, a new system that helps large language models solve problems better by focusing on specific tricky steps rather than general task summaries. Unlike earlier methods that give broad advice for whole tasks, HippoSpark looks at the exact moment where the model gets stuck and retrieves tailored experience to guide it. They tested this approach on math, science, and programming tasks and found it works better than previous techniques. The authors suggest that giving focused, state-level guidance is more effective than generic help for complex reasoning.
large language modelsexperience distillationstate-level retrievalcomplex reasoningtask-level promptingbottlenecksmathematical benchmarksscientific benchmarksprogramming benchmarks
Authors
Jingyao Liu, Danling Meng, Chen Huang, Yukun Yan, Zhenghao Liu, Wenqiang Lei, See-Kiong Ng, Maosong Sun
Abstract
Distilling historical trajectories into reusable experience to enhance future problem-solving has become a focal point of recent LLM research. However, existing methods predominantly operate at the task level, leveraging general summaries or rules under the assumption that analogous tasks share universal solution patterns. This approach often fails in complex reasoning, which typically falters at local bottlenecks that require precise, state-specific guidance rather than broad heuristics. We introduce HippoSpark, a state-level experience system that performs on-demand retrieval tailored to the immediate needs of the current reasoning state. Across mathematical, scientific, and programming benchmarks, HippoSpark consistently outperforms both standard prompting and task-level experience baselines. Our findings reveal that the most effective experience systems are those that provide actionable guidance at critical bottlenecks rather than serving as generic task-level context. Our code is available at https://github.com/DanlingMeng/HippoSpark.