arXiv：cs.CL· Bin Zheng, Zheng Chen·· 16 小时前
面向类型化决策模型的选择性推理：何时开始、读取与停止
Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop
AI 导读
研究团队在冻结的 Qwen3.5-4B 主干上训练了一个共享候选答案读取器，用收益门控启动推理、自然状态触发显式读取、熵规则接受答案并终止生成，从而统一直接决策与按需可中断推理。
正文
Abstract:Typed decision models need an interface that can use additional reasoning while retaining candidate outputs. We train a shared candidate-answer reader on a frozen Qwen3.5-4B backbone to unify direct decisions with on-demand, interruptible reasoning. The reader returns distributions from empty and partial reasoning prefixes. A benefit gate starts reasoning, a natural-state trigger requests explicit reads, and an entropy rule accepts acquired answers and cancels generation. Measurements distinguish answer-query work, generated output, and pipeline time. Across 768 questions from ProofWriter, LogiQA 2.0, and ARC-Challenge, accuracy rises from 70.96% for direct reading to 84.38%. Relative to 512-token terminal reading, accuracy improves by 7.29 percentage points with 2.25% more mean output work. Relative to 1,024-token terminal reading, pipeline time falls by 46.30% with an accuracy difference of -1.17 points. Exact-prefix controls isolate the saving from interruption, while read-scheduling controls distinguish reduced observation work from shorter response time.
| Comments: | 24 pages, 4 figures. Both authors contributed equally. Code and evaluation materials: this https URL |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.19644 [cs.CL] |
| (or arXiv:2610.19644v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19644 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zheng Chen [view email]
[v1]
Wed, 14 Oct 2026 06:13:42 UTC (303 KB)
来源：arXiv：cs.CL · arxiv.org