arXiv cs.CL· Thanni Adewuyi, Adepoju Jeremiah Moyondafoluwa, Dorcas Amapu, Dieko Whenu, Sylvia Essessien, Sandra Oko-Oba, Pelumi Onafadeji, Lily Aimas-Love, Samuel Omolaiye, Abiodun Adereni·· 2 天前AI 评分24
MamaReason:临床语言模型中的未完成推理与答案恢复
MamaReason: Unfinished Reasoning and Answer Recovery in Clinical Language Models
AI 导读
基于 Gemma 4 31B 微调的临床语言模型 MamaReason 在 MamaBench 的 217 组母婴健康病例上表现出“答案延续”现象:显式承诺提取器在 94 次失败的 RL 首轮推理中识别出 75 次诊断,在 93 次失败的 SFT 首轮中同样识别出 75 次。
正文 · 原文
Abstract:Answer continuation often returns a diagnosis already stated in unfinished clinical reasoning. We study this behavior in MamaReason, adapted from dense Gemma 4 31B, on MamaBench's 217 paired maternal and child-health presentations. An explicit-commitment extractor identifies a diagnosis in 75 of 94 failed RL first passes and 75 of 93 failed SFT first passes. The continuation matches that diagnosis in 67/75 RL cases and 58/75 SFT cases, using diagnosis strings alone. Reading these commitments without additional generation recovers 30 of the 40 RL pairs restored by continuation under the archived correctness labels. Recurrence controls distinguish routine mentions from repeated answer selection. An offline second-commitment trigger measures the tradeoff between earlier readout and paired accuracy. The resulting decomposition separates diagnostic content already present in a trace, completion of an accepted answer, and paired diagnostic correctness. For clinical language-model evaluation, these are distinct outcomes that become visible when the same reasoning trace is examined through multiple readouts.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.18264 [cs.CL] |
| (or arXiv:2611.18264v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.18264 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Thanni Adewuyi Ayomide [view email]
[v1]
Fri, 13 Nov 2026 14:58:11 UTC (35 KB)
来源:arXiv cs.CL · arxiv.org