Create eval.yaml
#3
by madhavan113 - opened
eval.yaml
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: APEX
|
| 2 |
+
description: >
|
| 3 |
+
APEX assesses whether frontier models are capable of performing
|
| 4 |
+
economically valuable tasks across four jobs: investment banking
|
| 5 |
+
associate, management consultant, big law associate, and primary
|
| 6 |
+
care physician (MD). 400 test cases (100 per job), graded using
|
| 7 |
+
a Judge LM. Created by experts from Goldman Sachs, McKinsey,
|
| 8 |
+
Latham & Watkins, and leading medical institutions.
|
| 9 |
+
evaluation_framework: apex-evals
|
| 10 |
+
tasks:
|
| 11 |
+
- id: apex-v1
|
| 12 |
+
config: default
|
| 13 |
+
split: train
|