arXiv cs.AI· Taebong Kim, Youngsik Hong, Minsik Kim, Sunyoung Choi, Jaewon Jang, Junghoon Shin, Minseo Kim·· 11 小时前AI 评分17
Darwin Family：无需训练的 MRI-Trust 加权进化合并，提升语言模型推理能力
Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Scaling of Language-Model Reasoning
AI 导读
Darwin Family 提出一种无需梯度训练的进化式模型合并框架，通过无梯度权重空间重组复用已有 checkpoint 中的潜在能力。其旗舰模型 Darwin-27B-Opus 在 GPQA Diamond 上达到 86.9%，在 1,252 个受评模型中排名第 6，超过其完全训练的基础模型。
正文 · 原文
This paper has been withdrawn by arXiv Admin
No PDF available, click to view other formats
Abstract:We present Darwin Family, a framework for training-free evolutionary merging of large language models via gradient-free weight-space recombination. We ask whether frontier-level reasoning performance can be improved without additional training, by reorganizing latent capabilities already encoded in existing checkpoints. Darwin introduces three key ideas: (i) a 14-dimensional adaptive merge genome enabling fine-grained component- and block-level recombination; (ii) MRI-Trust Fusion, which adaptively balances diagnostic layer-importance signals with evolutionary search through a learnable trust parameter; and (iii) an Architecture Mapper that enables cross-architecture breeding between heterogeneous model families. Empirically, the flagship Darwin-27B-Opus achieves 86.9% on GPQA Diamond, ranking #6 among 1,252 evaluated models, and outperforming its fully trained foundation model without any gradient-based training. Across scales from 4B to 35B parameters, Darwin models consistently improve over their parents, support recursive multi-generation evolution, and enable a training-free evolutionary merge that combines Transformer- and Mamba-based components. Together, the Darwin Family demonstrates that diagnostic-guided evolutionary merging is a practical and reproducible alternative to costly post-training pipelines for reasoning-centric language models.
| Comments: | arXiv admin note: This submission has been withdrawn because it does not meet arXiv's research content quality standards |
| Subjects: | Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.14386 [cs.NE] |
| (or arXiv:2605.14386v2 [cs.NE] for this version) | |
| https://doi.org/10.48550/arXiv.2605.14386 arXiv-issued DOI via DataCite |
Submission history
From: arXiv Admin [view email]
[v1]
Thu, 14 May 2026 05:09:12 UTC (520 KB)
[v2]
Tue, 3 Nov 2026 22:38:53 UTC (1 KB) (withdrawn)
来源：arXiv cs.AI · arxiv.org