New Method Merges Pre-trained Models Using Permutation and Reciprocal Orthogonality
A new arXiv paper proposes merging pre-trained models via permutation-based alignment, building on 2026 findings that wide, independently trained networks can show linear mode connectivity with softmax temperature.
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-11-11 05:00 · arXiv cs.LG
Permutation-Based Pre-trained Model Merging via Reciprocal Orthogonality