Qwen3.5-9B · valence set-point +5 SD (merged)

joshycodes/Qwen3.5-9B-valence-setpoint-plus5-lora merged into the full Qwen3.5-9B checkpoint (identical layout: Qwen3_5ForConditionalGeneration, vision tower and MTP weights untouched), so it can be served by stock vLLM like the base model. See the adapter card for the method, the valence axis and results.

Chat template: thinking is off by default (the adapter was trained and evaluated in non-thinking mode); pass enable_thinking=true to turn it on. Otherwise the template is Qwen3.5's own.

Research artifact; not intended for deployment.

Downloads last month
13
Safetensors
Model size
10B params
Tensor type
BF16
·
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for joshycodes/Qwen3.5-9B-valence-setpoint-plus5

Finetuned
Qwen/Qwen3.5-9B
Finetuned
(983)
this model