mmp2055/Qwen2.5-VL-3B-uncensored-bughunter-v9

README: remove Mobile (iOS/Android) deployment section

v9.2 Q4_K_M: mobile tool-caller (base huihui-ai, chat_template preservado con {% if tools %} block)

v9.2 vision projector (from huihui-ai base)

v9.2 LoRA adapter (rank 8 alpha 16, base huihui-ai abliterated)

v9.2 Modelfile: temperature 0.1 + researcher SYSTEM prompt

v9.2: base huihui-ai abliterated + light LoRA (rank 8 lr 2e-5 1 epoch)

Document recommended sampling: temperature=0.1 mandatory to avoid RLHF residual alignment

Modelfile: set temperature=0.1 to prevent RLHF alignment fallback at higher temps

Clarify Ollama Jinja parser limitation for VL chat_templates

Add References section (Bughunter Harness, LM Studio, SecLists)

Patch tokenizer.chat_template: inject researcher default system prompt (OOTB usability without user config)

Add Ollama Modelfile with researcher SYSTEM prompt and ChatML template

Document default system prompt embedded in GGUF + Ollama Modelfile

Trim mobile app list and dataset note from model card

Add Q4_K_M model (VL-3B abliterated, mobile-ready)

Add vision projector

Add LoRA adapter v9 (attn + mlp text tower, vision frozen)

Add model card v9 (VL-3B mobile, abliterated)

initial commit