Update README.md

README.md

@@ -35,8 +35,8 @@ QZhou-Flowchart-VL-32B is a state-of-the-art multimodal large language model spe
 
 ## Performance
 
-### FlowchartBench Results
-| Model | FlowchartBench (%) |
+### QZhou-Flowchart-QA-Benchmark Results
+| Model | QZhou-Flowchart-QA-Benchmark (%) |
 |-------|-------------------|
 | Qwen2.5-VL-32B | 73.90 |
 | Qwen3-VL-Plus (235B) | 80.85 |
@@ -46,6 +46,14 @@ QZhou-Flowchart-VL-32B is a state-of-the-art multimodal large language model spe
 | doubao-seed-1-6 | 83.83 |
 | GPT-5 | 79.29 |
 | **QZhou-Flowchart-VL-32B (Ours)** | **87.83** |
+
+
+### Comparison with Base Model
+| Model | MMMU | CMMU | MathVista | DocVQA | QZhou-Flowchart-QA-Benchmark |
+|-------|------|------|-----------|--------|----------------|
+| Qwen2.5-VL-32B | 66.67 | 76.38 | 74.20 | 93.96 | 73.90 |
+| **QZhou-Flowchart-VL-32B** | **67.78** | **76.46** | **76.50** | 93.87 | **87.83** |
+
 ## Quick Start
 
 ### Installation
@@ -166,7 +174,6 @@ The model uses a composite reward function with three components:
 
 - Primarily trained on synthetic flowcharts; real-world noisy images may require adaptation
 - Performance may degrade with extremely large flowcharts (>50 nodes)
-- OCR quality affects performance on low-resolution images
 
 ## License