arXiv：cs.CL· Mingyeong Kim, Gyusang Cho, Juho Lee, Giung Nam·· 16 小时前
通过流匹配实现理由分布蒸馏的分布式连续隐空间推理
Rationale Distribution Distillation via Flow Matching for Distributional Continuous Latent Reasoning
AI 导读
研究者提出理由分布蒸馏（RDD），通过从多条文本推理路径蒸馏，在定长隐序列上训练条件流，并用 within-query mixup 插值同一查询的完整隐理由以扩展蒸馏目标。该流非自回归地生成整段隐序列，再由冻结的语言模型从采样理由上下文自回归解码答案，在高等数学任务上取得了比所评估的点值隐推理基线更高的答案覆盖率。
正文
Abstract:Sampling multiple chain-of-thought rationales from a language model can uncover correct solutions missed by a single attempt. However, these rationales are typically lengthy, making repeated autoregressive generation costly. Continuous latent reasoning offers a compact alternative, but point-valued methods condition repeated answer sampling on a single rationale, which can limit exploration of diverse paths and answer coverage. We thus propose rationale distribution distillation (RDD), training a conditional flow over fixed-length latent sequences by distilling from multiple textual reasoning paths. Furthermore, we expand the distillation targets through within-query mixup, interpolating between complete latent rationales for the same query. The flow nonautoregressively generates the entire latent sequence, and a frozen language model autoregressively decodes answers from sampled rationale contexts. Empirically, the resulting distributional reasoner achieves higher answer coverage than the evaluated point-valued latent reasoning baselines, particularly on advanced mathematical tasks.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.19871 [cs.CL] |
| (or arXiv:2610.19871v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19871 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Giung Nam [view email]
[v1]
Wed, 14 Oct 2026 09:30:04 UTC (441 KB)
来源：arXiv：cs.CL · arxiv.org