[diffusion] fix: align negative prompt with official readme for new model (#16222)

python/sglang/multimodal_gen/configs/sample/qwenimage.py

@@ -15,6 +15,13 @@ class QwenImageSamplingParams(SamplingParams):
     num_inference_steps: int = 50
 
 
+@dataclass
+class QwenImage2512SamplingParams(QwenImageSamplingParams):
+    negative_prompt: str = (
+        "低分辨率，低画质，肢体畸形，手指畸形，画面过饱和，蜡像感，人脸无细节，过度光滑，画面具有AI感。构图混乱。文字模糊，扭曲。"
+    )
+
+
 @dataclass
 class QwenImageEditPlusSamplingParams(QwenImageSamplingParams):
     # Denoising stage