Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.11625 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911673215877120 |
|---|---|
| author | Zhou, Zhaomeng Zhang, Lan Wang, Junyang Yuan, Mu Lin, Junda |
| author_facet | Zhou, Zhaomeng Zhang, Lan Wang, Junyang Yuan, Mu Lin, Junda |
| contents | Large reasoning models (LRMs) improve problem solving through extended reasoning, but often misallocate test-time compute. Existing efficiency methods reduce cost by compressing reasoning traces or conditioning budget on perceived difficulty, yet largely overlook solvability. As a result, they may spend large budgets on queries beyond the model's capability while compressing hard-but-solvable queries that require deeper reasoning. In this work, we formulate adaptive reasoning as a computational investment under uncertainty, where budget should follow the expected return of reasoning rather than perceived difficulty alone. To instantiate this principle, we propose Budget-Efficient Thinking (BET), a two-stage framework that combines behavioral cold-start with GRPO under an investment-cost-aware reward. By aligning solve-or-fold decisions with rollout-derived solvability, BET learns three behaviors: (1) short solve, answering easy queries concisely; (2) nice fold, abstaining early when continued reasoning has near-zero expected return; and (3) hero call, preserving sufficient compute for hard-but-solvable queries. Across seven benchmarks and three base models, BET reduces reasoning tokens by ~55% on average while achieving overall performance improvements, and transfers zero-shot from mathematical reasoning to scientific QA and logical reasoning with comparable efficiency gains. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_11625 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Nice Fold or Hero Call: Learning Budget-Efficient Thinking for Adaptive Reasoning Zhou, Zhaomeng Zhang, Lan Wang, Junyang Yuan, Mu Lin, Junda Artificial Intelligence Large reasoning models (LRMs) improve problem solving through extended reasoning, but often misallocate test-time compute. Existing efficiency methods reduce cost by compressing reasoning traces or conditioning budget on perceived difficulty, yet largely overlook solvability. As a result, they may spend large budgets on queries beyond the model's capability while compressing hard-but-solvable queries that require deeper reasoning. In this work, we formulate adaptive reasoning as a computational investment under uncertainty, where budget should follow the expected return of reasoning rather than perceived difficulty alone. To instantiate this principle, we propose Budget-Efficient Thinking (BET), a two-stage framework that combines behavioral cold-start with GRPO under an investment-cost-aware reward. By aligning solve-or-fold decisions with rollout-derived solvability, BET learns three behaviors: (1) short solve, answering easy queries concisely; (2) nice fold, abstaining early when continued reasoning has near-zero expected return; and (3) hero call, preserving sufficient compute for hard-but-solvable queries. Across seven benchmarks and three base models, BET reduces reasoning tokens by ~55% on average while achieving overall performance improvements, and transfers zero-shot from mathematical reasoning to scientific QA and logical reasoning with comparable efficiency gains. |
| title | Nice Fold or Hero Call: Learning Budget-Efficient Thinking for Adaptive Reasoning |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2605.11625 |