exam-v1: Codex 19/76 vs Trace 74/76 (not Hub gate)
Browse files
README.md
CHANGED
|
@@ -117,9 +117,21 @@ Weights learn **how to screen**. Facts that move (SOP revs, notices) stay in the
|
|
| 117 |
|
| 118 |
## exam-v1 (not the Hub gate)
|
| 119 |
|
| 120 |
-
|
| 121 |
|
| 122 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 123 |
|
| 124 |
## Quickstart
|
| 125 |
|
|
|
|
| 117 |
|
| 118 |
## exam-v1 (not the Hub gate)
|
| 119 |
|
| 120 |
+
Same 76 files. Same grader. Frontier seats are naked JSON. Trace is the product seat. Holdout n=12 is the Hub number.
|
| 121 |
|
| 122 |
+
|
| 123 |
+
|
| 124 |
+
|
| 125 |
+
|
| 126 |
+
| Seat | All-pass | False kill | SOP clock |
|
| 127 |
+
|---|---|---|---|
|
| 128 |
+
| Qwen 2.5 7B | 0.00 (0/76) | 0.12 | 0.00 |
|
| 129 |
+
| Qwen + dated RAG | 0.16 (12/76) | 0.04 | 0.70 |
|
| 130 |
+
| Qwen + RAG + tools | 0.20 (15/76) | 0.03 | 0.71 |
|
| 131 |
+
| Codex | 0.25 (19/76) | 0 | 0.89 |
|
| 132 |
+
| **Trace** | **0.97 (74/76)** | **0** | **1.00** |
|
| 133 |
+
|
| 134 |
+
Codex occupancy 8/8; pass family 0/20 (labels `needs_facts`, not a false kill). Trace affiliation 6/8; every other family 8/8 or 20/20. Claude and Grok exam still running. `verified: false`.
|
| 135 |
|
| 136 |
## Quickstart
|
| 137 |
|