Spaces:
Running
Running
Deploy eval-card-registry service
Browse files
packages/eval-entity-resolver/src/eval_entity_resolver/eee.py
CHANGED
|
@@ -101,6 +101,10 @@ _METRIC_KEYWORDS: list[tuple[str, str]] = [
|
|
| 101 |
(r"discrete[\s_-]*win[\s_-]*rate", "Discrete Win Rate"),
|
| 102 |
(r"win[\s_-]*rate", "Win Rate"),
|
| 103 |
(r"mean[\s_-]*response[\s_-]*time", "Mean Response Time"),
|
|
|
|
|
|
|
|
|
|
|
|
|
| 104 |
(r"mean[\s_-]*score", "Mean Score"),
|
| 105 |
# HELM's near-miss exact-match variants are separately computed stats.
|
| 106 |
(r"prefix[\s_-]*quasi[\s_-]*exact[\s_-]*match", "Prefix Quasi-Exact Match"),
|
|
|
|
| 101 |
(r"discrete[\s_-]*win[\s_-]*rate", "Discrete Win Rate"),
|
| 102 |
(r"win[\s_-]*rate", "Win Rate"),
|
| 103 |
(r"mean[\s_-]*response[\s_-]*time", "Mean Response Time"),
|
| 104 |
+
# "Mean Progress Score" is a distinct measurement (fraction of a
|
| 105 |
+
# scenario's steps completed), not a mean of scores — it must be caught
|
| 106 |
+
# before the generic `score` pattern swallows it.
|
| 107 |
+
(r"mean[\s_-]*progress[\s_-]*score", "Mean Progress Score"),
|
| 108 |
(r"mean[\s_-]*score", "Mean Score"),
|
| 109 |
# HELM's near-miss exact-match variants are separately computed stats.
|
| 110 |
(r"prefix[\s_-]*quasi[\s_-]*exact[\s_-]*match", "Prefix Quasi-Exact Match"),
|