# 微调反转的引力解释

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-06-26 08:00
- AIHOT 分数：51
- AIHOT 链接：https://aihot.virxact.com/items/cmr0svezl04s3slolki0kj19a
- 原文链接：https://arxiv.org/abs/2606.28525

## AI 摘要

针对无害数据微调可部分撤销早期训练获得的安全行为，论文提出几何假设：早期训练创建主导行为流形，后续对齐仅产生浅层位移，后续微调会继承指向主导流形的持久反转分量v_rev。实验显示表征沿v_rev的对齐从首次更新后cos=0.429±0.052升至第20步0.647±0.021，24个运行-步对均超过各向同性零假设p99。选择性阻止沿v_rev运动，使最终对齐从0.648±0.009降至-0.211±0.021，harmfulness从19.0%±4.0%降至8.5%±1.5%，任务成本极小，表明v_rev是早期对齐后反转的因果中介。

## 正文

Fine-tuning on harmless data can partially undo behaviors acquired earlier in training. Safety can erode under benign post-alignment updates, unlearned capabilities can re-emerge, latent traits can transfer through apparently unrelated supervision, and related post-alignment fragility appears in other generative settings. We argue these phenomena are usefully viewed through a common training-history lens. Our hypothesis is geometric: large early training phases create dominant behavioral manifolds, while later alignment or specialization phases are shallower displacements from them. Subsequent fine-tuning can therefore inherit a persistent reversion component pointing back toward a witness of the dominant manifold. We call this the gravitational interpretation of fine-tuning reversion. Across our main settings, representational drift rapidly acquires a component along a history-defined reversion direction (v_rev). In our main track, alignment with v_rev rises from cos = 0.429 +/- 0.052 after the first update to 0.647 +/- 0.021 by step 20. Across 24 run-step pairs, every observed alignment exceeds the p99 of an isotropic activation-space null. We demonstrate that selectively blocking motion along v_rev changes the final alignment at T=100 from 0.648 +/- 0.009 to -0.211 +/- 0.021 and reduces harmfulness from 19.0% +/- 4.0% to 8.5% +/- 1.5% with little task cost. These results support v_rev as a causally relevant mediator of early post-alignment reversion in our setup. Importantly, we do not claim that v_rev is the unique safety direction, nor that the dominant manifold is directly observed; rather, we identify a robust, history-defined direction that explains and partially controls early reversion dynamics.