NCSOFT/VARCO-VISION-2.0-14B
Full fine-tune (fewer layers) vs Qwen/Qwen3-14B. Qwen/Qwen3-14B matches part of NCSOFT/VARCO-VISION-2.0-14B (441 of its 869 tensors; the rest of NCSOFT/VARCO-VISION-2.0-14B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 40 of 40 layers.
Tensors changed: Norms 161/161, MLP 120/120, Attention 160/160
1 other repos hold byte-for-byte identical weight files.