NousResearch/atropos
1
0
Fork 0
mirror of https://github.com/NousResearch/atropos.git synced 2026-04-19 12:57:58 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

[pre-commit.ci] auto fixes from pre-commit.com hooks

Browse source 
for more information, see https://pre-commit.ci
This commit is contained in:
pre-commit-ci[bot] 2026-01-12 23:29:15 +00:00
parent 22884d2bf7
commit d84e3c70b7
16 changed files with 270 additions and 143 deletions
Download patch file Download diff file Expand all files Collapse all files

18
environments/eval_environments/hallusionbench_environment.py
View file

@ -36,7 +36,9 @@ class HallusionBench(EvalBase):
except Exception:
pass
if all_data:
print(f"Loaded {len(all_data)} examples from HallusionBench (combined)")
print(
f"Loaded {len(all_data)} examples from HallusionBench (combined)"
)
return all_data
raise ValueError(f"Could not load HallusionBench dataset: {e}")
except Exception:
@ -62,10 +64,12 @@ class HallusionBench(EvalBase):
content = []
if image_base64:
content.append({
"type": "image_url",
"image_url": {"url": f"data:image/png;base64,{image_base64}"},
})
content.append(
{
"type": "image_url",
"image_url": {"url": f"data:image/png;base64,{image_base64}"},
}
)
content.append({"type": "text", "text": prompt})
return [{"role": "user", "content": content}]
@ -79,8 +83,8 @@ class HallusionBench(EvalBase):
if response_lower.startswith("no"):
return "No"
yes_patterns = [r'\byes\b', r'\btrue\b', r'\bcorrect\b']
no_patterns = [r'\bno\b', r'\bfalse\b', r'\bincorrect\b']
yes_patterns = [r"\byes\b", r"\btrue\b", r"\bcorrect\b"]
no_patterns = [r"\bno\b", r"\bfalse\b", r"\bincorrect\b"]
for pattern in yes_patterns:
if re.search(pattern, response_lower):