Fin-SD 如何提升金融推理:回溯自蒸餾的適用條件
How Much Does the Hint Matter? Scope Conditions for Hindsight Self-Distillation in Financial Reasoning
Fin-SD 透過學生模型自行撰寫錯誤評論,並以截斷至錯誤前的正確推理前綴進行按策略自蒸餾,改善多步金融推理,無須外部教師模型提供梯度。在 FinanceReasoning Hard 子集上,Fin-SD 比基礎模型準確率高 3.03 個百分點;以完整參考解答進行按策略自蒸餾時,準確率相若,但生成的 tokens 少 9.5%。使用 GPT-OSS-120B 撰寫提示,效益亦未見明顯高於自行撰寫。
Published on Oct 26
Authors:
Abstract
Fin-SD improves local multi-step financial reasoning by using self-generated error critiques and truncated correct prefixes for on-policy distillation without external teachers.
Financial institutions operating under data-residency constraints are restricted to open-weight models small enough to locally host, which trail frontier systems on multi-step quantitative reasoning. On-policy distillation narrows such gaps by supervising a student on its own rollouts, but traditional OPD requires a strong external teacher, and on-policy self-distillation (OPSD) conditions on a full reference solution. We ask how much of that signal is necessary. We present Fin-SD, a localized hindsight self-distillation recipe for multi-step reasoning. Given a problem the student fails, the student itself identifies the first incorrect reasoning step and writes a short critique of that step. We truncate the student's trajectory immediately before the error (aiming to obtain a correct prefix), condition a frozen copy of the student on the same prefix plus the hint, and distill that hint-conditioned teacher into the student over a short rollout under reverse KL. No external model contributes a gradient. We apply Fin-SD to GPT-OSS-20B on a custom multi-step financial reasoning benchmark that requires deriving formula inputs from realistic scenarios via accounting conventions. We find: (i) replacing the critique with a generic-feedback control matches Fin-SD's entire 3.03 pp accuracy gain over the base model on the FinanceReasoning Hard subset, suggesting the gain comes from where supervision is applied, not hint content; (ii) the model writing the hint does not need to be stronger than the model receiving it: hints from same-family GPT-OSS-120B, at roughly 6x the parameters, yield no materially greater benefits than self-written hints; and (iii) Fin-SD matches on-policy self-distillation from full reference solutions on accuracy while generating 9.5% fewer tokens. With Fin-SD, general reasoning on MMLU-Pro and a fourteen-metric compliance audit on FinTrust show no material regression.
View arXiv page View PDF GitHub Add to collection
Get this paper in your agent:
hf papers read 2610.36082
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash
Models citing this paper 0
No model linking this paper
Cite arxiv.org/abs/2610.36082 in a model README.md to link it from this page.
Datasets citing this paper 0
No dataset linking this paper
Cite arxiv.org/abs/2610.36082 in a dataset README.md to link it from this page.
Spaces citing this paper 0
No Space linking this paper
Cite arxiv.org/abs/2610.36082 in a Space README.md to link it from this page.
Collections including this paper 0
No Collection including this paper
Add this paper to a collection to link it from this page.
來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co