seed: gpt-5.5.json
Browse files- tables/gpt-5.5.json +34 -0
tables/gpt-5.5.json
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "gpt-5.5",
|
| 3 |
+
"provider": "openai",
|
| 4 |
+
"capabilities": [
|
| 5 |
+
"coding",
|
| 6 |
+
"creative_synthesis",
|
| 7 |
+
"instruction_following",
|
| 8 |
+
"math_reasoning",
|
| 9 |
+
"planning_agentic",
|
| 10 |
+
"world_knowledge"
|
| 11 |
+
],
|
| 12 |
+
"skill_vector": [
|
| 13 |
+
0.89,
|
| 14 |
+
0.65,
|
| 15 |
+
0.71,
|
| 16 |
+
0.92,
|
| 17 |
+
0.83,
|
| 18 |
+
0.94
|
| 19 |
+
],
|
| 20 |
+
"source": "benchmark",
|
| 21 |
+
"confidence": [
|
| 22 |
+
"high",
|
| 23 |
+
"low",
|
| 24 |
+
"medium",
|
| 25 |
+
"high",
|
| 26 |
+
"medium",
|
| 27 |
+
"high"
|
| 28 |
+
],
|
| 29 |
+
"imputed_capabilities": [],
|
| 30 |
+
"support": null,
|
| 31 |
+
"subset_hash": null,
|
| 32 |
+
"date": "2026-06-11",
|
| 33 |
+
"notes": "Derived from published AI-lab benchmarks (SWE-bench, AIME, GPQA, MMLU-Pro, tau-bench, etc.), normalized to [0,1] and mapped onto the 6 capabilities. Cold-start prior: a `measured` extraction on the frozen probe set overrides this. See README.md."
|
| 34 |
+
}
|