Update README.md
Browse files
README.md
CHANGED
|
@@ -56,6 +56,13 @@ The model adopts an MoE language model, a native-resolution visual encoder (Moon
|
|
| 56 |
|
| 57 |
</div>
|
| 58 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
## Performance
|
| 60 |
|
| 61 |
With effective long-thinking abilitites, Kimi-VL-A3B-Thinking can match the performance of 30B/70B frontier open-source VLMs on MathVision benchmark:
|
|
|
|
| 56 |
|
| 57 |
</div>
|
| 58 |
|
| 59 |
+
> [!Note]
|
| 60 |
+
> Recommended parameter settings:
|
| 61 |
+
> - For **Thinking models**, it is recommended to use `Temperature = 0.6`.
|
| 62 |
+
> - For **Instruct models**, it is recommended to use `Temperature = 0.2`.
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
|
| 66 |
## Performance
|
| 67 |
|
| 68 |
With effective long-thinking abilitites, Kimi-VL-A3B-Thinking can match the performance of 30B/70B frontier open-source VLMs on MathVision benchmark:
|