Do Not Waste Your Rollouts: Recycling Search Experience for Efficient Test-Time Scaling
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911645475799040 |
|---|---|
| author | Wang, Xinglin Shi, Jiayi Feng, Shaoxiong Yuan, Peiwen Li, Yiwei Zhang, Yueqi Tan, Chuyi Zhang, Ji Pan, Boyuan Hu, Yao Li, Kan |
| author_facet | Wang, Xinglin Shi, Jiayi Feng, Shaoxiong Yuan, Peiwen Li, Yiwei Zhang, Yueqi Tan, Chuyi Zhang, Ji Pan, Boyuan Hu, Yao Li, Kan |
| contents | Test-Time Scaling enhances the reasoning capabilities of Large Language Models by allocating additional inference compute to broaden the exploration of the solution space. However, existing search strategies typically treat rollouts as disposable samples, where valuable intermediate insights are effectively discarded after each trial. This wasted rollout-level experience leads to substantial computational redundancy, as models repeatedly re-derive discovered conclusions and revisit known dead ends across extensive attempts. To bridge this gap, we propose \textbf{Recycling Search Experience (RSE)}, a self-guided, training-free strategy that turns test-time search from a series of isolated trials into a cumulative, experience-guided process. By actively distilling raw trajectories into a shared experience bank, RSE enables positive recycling of intermediate conclusions to shortcut redundant derivations and negative recycling of failure patterns to prune encountered dead ends. Theoretically, we provide an analysis that formalizes the efficiency gains of RSE over independent sampling in solving complex reasoning tasks. Empirically, extensive experiments on HMMT24, HMMT25, IMO-Bench, and HLE show that RSE consistently outperforms strong baselines under comparable computational budgets, establishing a strong compute-efficiency frontier for test-time scaling. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_21684 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Do Not Waste Your Rollouts: Recycling Search Experience for Efficient Test-Time Scaling Wang, Xinglin Shi, Jiayi Feng, Shaoxiong Yuan, Peiwen Li, Yiwei Zhang, Yueqi Tan, Chuyi Zhang, Ji Pan, Boyuan Hu, Yao Li, Kan Computation and Language Machine Learning Test-Time Scaling enhances the reasoning capabilities of Large Language Models by allocating additional inference compute to broaden the exploration of the solution space. However, existing search strategies typically treat rollouts as disposable samples, where valuable intermediate insights are effectively discarded after each trial. This wasted rollout-level experience leads to substantial computational redundancy, as models repeatedly re-derive discovered conclusions and revisit known dead ends across extensive attempts. To bridge this gap, we propose \textbf{Recycling Search Experience (RSE)}, a self-guided, training-free strategy that turns test-time search from a series of isolated trials into a cumulative, experience-guided process. By actively distilling raw trajectories into a shared experience bank, RSE enables positive recycling of intermediate conclusions to shortcut redundant derivations and negative recycling of failure patterns to prune encountered dead ends. Theoretically, we provide an analysis that formalizes the efficiency gains of RSE over independent sampling in solving complex reasoning tasks. Empirically, extensive experiments on HMMT24, HMMT25, IMO-Bench, and HLE show that RSE consistently outperforms strong baselines under comparable computational budgets, establishing a strong compute-efficiency frontier for test-time scaling. |
| title | Do Not Waste Your Rollouts: Recycling Search Experience for Efficient Test-Time Scaling |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2601.21684 |