Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2507.02076 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913923620405248 |
|---|---|
| author | Alomrani, Mohammad Ali Zhang, Yingxue Li, Derek Sun, Qianyi Pal, Soumyasundar Zhang, Zhanguang Hu, Yaochen Ajwani, Rohan Deepak Valkanas, Antonios Karimi, Raika Cheng, Peng Wang, Yunzhou Liao, Pengyi Huang, Hanrui Wang, Bin Hao, Jianye Coates, Mark |
| author_facet | Alomrani, Mohammad Ali Zhang, Yingxue Li, Derek Sun, Qianyi Pal, Soumyasundar Zhang, Zhanguang Hu, Yaochen Ajwani, Rohan Deepak Valkanas, Antonios Karimi, Raika Cheng, Peng Wang, Yunzhou Liao, Pengyi Huang, Hanrui Wang, Bin Hao, Jianye Coates, Mark |
| contents | Large language models (LLMs) have rapidly progressed into general-purpose agents capable of solving a broad spectrum of tasks. However, current models remain inefficient at reasoning: they apply fixed inference-time compute regardless of task complexity, often overthinking simple problems while underthinking hard ones. This survey presents a comprehensive review of efficient test-time compute (TTC) strategies, which aim to improve the computational efficiency of LLM reasoning. We introduce a two-tiered taxonomy that distinguishes between L1-controllability, methods that operate under fixed compute budgets, and L2-adaptiveness, methods that dynamically scale inference based on input difficulty or model confidence. We benchmark leading proprietary LLMs across diverse datasets, highlighting critical trade-offs between reasoning performance and token usage. Compared to prior surveys on efficient reasoning, our review emphasizes the practical control, adaptability, and scalability of TTC methods. Finally, we discuss emerging trends such as hybrid thinking models and identify key challenges for future work towards making LLMs more computationally efficient, robust, and responsive to user constraints. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_02076 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs Alomrani, Mohammad Ali Zhang, Yingxue Li, Derek Sun, Qianyi Pal, Soumyasundar Zhang, Zhanguang Hu, Yaochen Ajwani, Rohan Deepak Valkanas, Antonios Karimi, Raika Cheng, Peng Wang, Yunzhou Liao, Pengyi Huang, Hanrui Wang, Bin Hao, Jianye Coates, Mark Artificial Intelligence Machine Learning Large language models (LLMs) have rapidly progressed into general-purpose agents capable of solving a broad spectrum of tasks. However, current models remain inefficient at reasoning: they apply fixed inference-time compute regardless of task complexity, often overthinking simple problems while underthinking hard ones. This survey presents a comprehensive review of efficient test-time compute (TTC) strategies, which aim to improve the computational efficiency of LLM reasoning. We introduce a two-tiered taxonomy that distinguishes between L1-controllability, methods that operate under fixed compute budgets, and L2-adaptiveness, methods that dynamically scale inference based on input difficulty or model confidence. We benchmark leading proprietary LLMs across diverse datasets, highlighting critical trade-offs between reasoning performance and token usage. Compared to prior surveys on efficient reasoning, our review emphasizes the practical control, adaptability, and scalability of TTC methods. Finally, we discuss emerging trends such as hybrid thinking models and identify key challenges for future work towards making LLMs more computationally efficient, robust, and responsive to user constraints. |
| title | Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs |
| topic | Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2507.02076 |