CodeGoat24/UnifiedReward-2.0-qwen3vl-2b
Partial fine-tune vs Qwen/Qwen3-VL-2B-Instruct. Partial change: 85% of the weights changed (attention, embeddings, mlp, norms), in 28 of 28 layers.
Tensors changed: Embeddings 2/4, Norms 111/217, MLP 96/196, Attention 112/208
3 derivatives checked
- mradermacher/UnifiedReward-2.0-qwen3vl-2b-i1-GGUF: Quantization (fewer layers)
- mradermacher/UnifiedReward-2.0-qwen3vl-2b-GGUF: Same weights, converted (fewer layers)
- CodeGoat24/UnifiedReward-Think-qwen3vl-2b: Partial fine-tune