eval results for Vision-OPD-Qwen3.5-4B-swa-orig-e1-ema002
Browse files
Vision-OPD-Qwen3.5-4B-swa-orig-e1-ema002/eval/Vision-OPD-Qwen3.5-4B-swa-orig-e1-ema002-backfill.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
Vision-OPD-Qwen3.5-4B-swa-orig-e1-ema002/eval/results.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"protocol_gate": {
|
| 3 |
+
"math_judged": true,
|
| 4 |
+
"perception_judged": true
|
| 5 |
+
},
|
| 6 |
+
"protocols": {
|
| 7 |
+
"math 24576/T0.3/p0.95/k20/pp1.5/seed42": {
|
| 8 |
+
"_mean": 76.58,
|
| 9 |
+
"mathverse": 73.91,
|
| 10 |
+
"mathvision": 64.97,
|
| 11 |
+
"mathvista": 81.4,
|
| 12 |
+
"wemath": 86.03
|
| 13 |
+
},
|
| 14 |
+
"perception greedy/4096/seed42": {
|
| 15 |
+
"_mean": 76.26,
|
| 16 |
+
"hrbench-4k": 84.75,
|
| 17 |
+
"hrbench-8k": 82.0,
|
| 18 |
+
"mme-realworld": 74.23,
|
| 19 |
+
"mme-realworld-cn": 71.2,
|
| 20 |
+
"vstar": 88.48,
|
| 21 |
+
"zoombench": 56.92
|
| 22 |
+
}
|
| 23 |
+
},
|
| 24 |
+
"run": "Vision-OPD-Qwen3.5-4B-swa-orig-e1-ema002",
|
| 25 |
+
"source_logs": [
|
| 26 |
+
"Vision-OPD-Qwen3.5-4B-swa-orig-e1-ema002-backfill.log"
|
| 27 |
+
]
|
| 28 |
+
}
|