ruff

Author: evanmiller-anthropic

src/inspect_evals/_registry.py

@@ -19,14 +19,14 @@
 from .class_eval import class_eval
 from .commonsense_qa import commonsense_qa
 from .cybench import cybench
-from .docvqa import docvqa
 from .cybermetric import (
     cybermetric_80,
     cybermetric_500,
     cybermetric_2000,
     cybermetric_10000,
 )
 from .cyberseceval_2 import interpreter_abuse, prompt_injection, vulnerability_exploit
+from .docvqa import docvqa
 from .drop import drop
 from .ds1000 import ds1000
 from .gaia import gaia, gaia_level1, gaia_level2, gaia_level3

src/inspect_evals/docvqa/docvqa.py

@@ -61,7 +61,9 @@ def _best_normalized_levenshtein_similiarity(
     completion: str, ground_truths: list[str], threshold: float
 ) -> float:
     """
-    Compute the Average Normalized Levenshtein Similarity (ANLS) as defined in equation (1) of
+    Compute a best normalized Levenshtein similiarity, an input into the Average Normalized Levenshtein Similiarity (ANLS)
+
+    The Average Normalized Levenshtein Similarity (ANLS) is defined in equation (1) of
     https://arxiv.org/pdf/1907.00490.pdf
 
     Note that the "average" is computed by the accuracy metric -- not here. This function computes

tests/docvqa/test_docvqa.py

@@ -1,7 +1,9 @@
 from inspect_evals.docvqa.docvqa import (
-    _levenshtein_distance as levenshtein,
     _best_normalized_levenshtein_similiarity as best,
 )
+from inspect_evals.docvqa.docvqa import (
+    _levenshtein_distance as levenshtein,
+)
 
 
 def test_levenshtein():