DreamFast/Qwen3.5-4B-Uncensored-HauhauCS-Aggressive-Safetensor-Benchmark
Partial fine-tune vs Qwen/Qwen3.5-4B. Partial change: 20% of the weights changed (attention, mlp, norms), in 28 of 32 layers.
Tensors changed: Embeddings 0/4, MLP 32/196, Attention 34/320, Norms 26/203
3 derivatives checked
- minas2025/Qwen-3.5-4B-RP-Finetune: Quantization
- minas2025/Qwen-3.5-4B-Conversational-Uncensored: Full fine-tune
- zzlat/tuningTEST: Quantization