inclusionAI/ArmorOCR

Full fine-tune vs Qwen/Qwen3-VL-8B-Instruct. Fine-tune: 93% of the weights changed, across 36 of 36 layers.

Tensors changed: Embeddings 3/4, Norms 189/261, MLP 226/232, Attention 252/252, lm_head 1/1

2 derivatives checked

• prithivMLmods/ArmorOCR-GGUF: Same weights, converted (fewer layers)
• inclusionAI/ArmorOCR-GGUF: Quantization (fewer layers)

On the Hub · Downloads on Model Pulse