ABACUS:基于冻结判断模型的多步推理框架
ABACUS: Program-Guided Multistep Reasoning over Frozen Judgment Models
AI 导读
ABACUS 框架通过生成式教师离线设计程序、冻结判断模型(Kev)在线解决局部语义问题,将局部决策组合成多步推理。在 BIG-Bench Hard 三个任务的 600 个留出问题上,三个程序达到 99.67% macro accuracy,远高于直接使用 Kev 的 64.00% 和 Qwen3.5-4B 的 68.80%。
正文 · 原文
Abstract:Judgment models answer semantic questions by scoring a finite set of candidate outcomes. Can reusable programs turn these local decisions into powerful multistep reasoning? We introduce \method, a framework that jointly constructs judgment schemas and composition programs for frozen judgment models. A generative teacher designs the procedure offline; the judgment model resolves local semantic questions online, and the saved program composes its decisions into an answer. We study three allocations of design effort between humans and the teacher, focusing on construction through a shared interface. We instantiate the framework with Kev. On three selected BIG-Bench Hard tasks, nine independently constructed programs solve all 450 executions of 150 development inputs. A separate deployment study freezes three programs after development-set reference-agreement refinement and evaluates them on 600 held-out questions. These programs achieve near-perfect 99.67\% macro accuracy, versus 64.00\% for direct Kev and 68.80\% for final-answer-only Qwen3.5-4B. They attain comparable observed accuracy to Qwen's unrestricted response policy (99.07\%) with an \textbf{8.43-fold online speedup}: 0.255 versus 2.150 seconds per question. These results show how reusable programs can substantially expand the reasoning capability of a compact judgment model while keeping its parameters fixed.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.25963 [cs.AI] |
| (or arXiv:2610.25963v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.25963 arXiv-issued DOI via DataCite |
Submission history
From: Zelin Zang [view email]
[v1]
Mon, 19 Oct 2026 07:57:48 UTC (314 KB)
[v2]
Tue, 17 Nov 2026 09:07:38 UTC (494 KB)
来源:arXiv cs.AI · arxiv.org