{"ID":2847320,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2511.14772","arxiv_id":"2511.14772","title":"Test-time Scaling of LLMs: A Survey from A Subproblem Structure Perspective","abstract":"With this paper, we survey techniques for improving the predictive accuracy of pretrained large language models by allocating additional compute at inference time. In categorizing test-time scaling methods, we place special emphasis on how a problem is decomposed into subproblems and on the topological organization of these subproblems whether sequential, parallel, or tree-structured. This perspective allows us to unify diverse approaches such as Chain-of-Thought, Branch-Solve-Merge, and Tree-of-Thought under a common lens. We further synthesize existing analyses of these techniques, highlighting their respective strengths and weaknesses, and conclude by outlining promising directions for future research","short_abstract":"With this paper, we survey techniques for improving the predictive accuracy of pretrained large language models by allocating additional compute at inference time. In categorizing test-time scaling methods, we place special emphasis on how a problem is decomposed into subproblems and on the topological organization of...","url_abs":"https://arxiv.org/abs/2511.14772","url_pdf":"https://arxiv.org/pdf/2511.14772v1","authors":"[\"Zhuoyi Yang\",\"Xu Guo\",\"Tong Zhang\",\"Huijuan Xu\",\"Boyang Li\"]","published":"2025-11-01T18:41:23Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.AI\"]","methods":"[\"Large Language Model\",\"Language Model\",\"Generative Adversarial Network\"]","has_code":false}