arXiv：cs.LG（机器学习，全量分类）· Luca Zhou, Emanuele Rodol\`a·· 15 小时前AI 评分54
arXiv 论文研究模型合并对涌现能力的影响
On Emergent Capabilities and Model Merging
AI 导读
arXiv 论文（arXiv:2609.24504）研究模型合并（model merging）对涌现能力的影响，在激活 oracle 和涌现失齐模型两个测试平台、三个模型族上得出三重结论。
正文
Abstract:Fine-tuned checkpoints and adapters now fill public repositories, and the most common operation applied to these artifacts is model merging: arithmetic on their weights that assembles capabilities cheaply. We ask what this operation does to emergent capabilities: behaviors an artifact carries that were never an explicit training target. Studying two independent testbeds (activation oracles and emergent-misaligned models) across three model families, we find that the answer is threefold. First, merging preserves an emergent capability that both parents carry: merging two misaligned checkpoints retains most of their broad misalignment across the whole mixing range. Second, merging cannot create an emergent capability that is superadditive in its parents: no weighted merge of two single-task oracles reaches the jointly-trained oracle's auditing ability. Third, when only one parent carries the capability, merging dilutes it faster than the trained capability that accompanies it: the gap is significant in most settings. In short, emergent behaviors of an artifact do not compose the way its trained capability does.
| Comments: | main paper has 8 pages, 5 figures, and 4 tables |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.24504 [cs.LG] |
| (or arXiv:2609.24504v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2609.24504 arXiv-issued DOI via DataCite |
Submission history
From: Luca Zhou [view email]
[v1]
Mon, 21 Sep 2026 12:44:35 UTC (212 KB)
[v2]
Thu, 1 Oct 2026 13:56:04 UTC (212 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org