Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat:fix-type-handling-in-metrics #5

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

feat:fix-type-handling-in-metrics #5

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
47 changes: 32 additions & 15 deletions ragchecker/computation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@ def evaluate_precision(result: RAGResult):
return
assert result.answer2response is not None
answer2response = to_bool(result.answer2response)
if len(answer2response) > 0:
if isinstance(answer2response, (np.ndarray, list)) and len(answer2response) > 0:
result.metrics[metrics.precision] = np.mean(answer2response)
else:
result.metrics[metrics.precision] = 0.
Expand All @@ -26,7 +26,7 @@ def evaluate_recall(result: RAGResult):
return
assert result.response2answer is not None
response2answer = to_bool(result.response2answer)
if len(response2answer) > 0:
if isinstance(response2answer, (np.ndarray, list)) and len(response2answer) > 0:
result.metrics[metrics.recall] = np.mean(response2answer)
else:
result.metrics[metrics.recall] = 0.
Expand Down Expand Up @@ -61,11 +61,17 @@ def evaluate_retrieval(result: RAGResult):
"""Evaluate retrieval metrics together as they share the same intermediate results."""
assert result.retrieved2answer is not None
retrieved2answer = to_bool(result.retrieved2answer)
if len(retrieved2answer) > 0 and len(retrieved2answer[0]) > 0:
claim_recalled = np.max(retrieved2answer, axis=1)
result.metrics[metrics.claim_recall] = np.mean(claim_recalled)
psg_useful = np.max(retrieved2answer, axis=0)
result.metrics[metrics.context_precision] = np.mean(psg_useful)
if isinstance(retrieved2answer, (np.ndarray, list)) and len(retrieved2answer) > 0:
if isinstance(retrieved2answer[0], (np.ndarray, list)) and len(retrieved2answer[0]) > 0:
claim_recalled = np.max(retrieved2answer, axis=1)
result.metrics[metrics.claim_recall] = np.mean(claim_recalled)
psg_useful = np.max(retrieved2answer, axis=0)
result.metrics[metrics.context_precision] = np.mean(psg_useful)
else:
# This handles the case where retrieved2answer is a 1D array of booleans
claim_recalled = retrieved2answer
result.metrics[metrics.claim_recall] = np.mean(claim_recalled)
result.metrics[metrics.context_precision] = 0.
else:
result.metrics[metrics.claim_recall] = 0.
result.metrics[metrics.context_precision] = 0.
Expand All @@ -77,11 +83,16 @@ def evaluate_context_utilization(result: RAGResult):
assert result.retrieved2answer is not None and result.response2answer is not None
retrieved2answer = to_bool(result.retrieved2answer)
response2answer = to_bool(result.response2answer)
if len(retrieved2answer) > 0 and len(retrieved2answer[0]) > 0:
claim_recalled = np.max(retrieved2answer, axis=1)
if np.sum(claim_recalled) > 0:
claim_used = claim_recalled & response2answer
result.metrics[metrics.context_utilization] = np.sum(claim_used) / np.sum(claim_recalled)

if isinstance(retrieved2answer, (np.ndarray, list)) and len(retrieved2answer) > 0:
# Check if retrieved2answer is a 1D array of booleans
if np.ndim(retrieved2answer) == 1 or (np.ndim(retrieved2answer) > 1 and len(retrieved2answer[0]) > 0):
claim_recalled = np.max(retrieved2answer, axis=1) if np.ndim(retrieved2answer) > 1 else retrieved2answer
if np.sum(claim_recalled) > 0:
claim_used = claim_recalled & response2answer
result.metrics[metrics.context_utilization] = np.sum(claim_used) / np.sum(claim_recalled)
else:
result.metrics[metrics.context_utilization] = 0.
else:
result.metrics[metrics.context_utilization] = 0.
else:
Expand All @@ -107,7 +118,9 @@ def evaluate_noise_sensitivity(result: RAGResult):
retrieved2response = to_bool(result.retrieved2response)
answer2response = to_bool(result.answer2response)
retrieved2answer = to_bool(result.retrieved2answer)
if len(answer2response) > 0 and len(retrieved2response[0]) > 0 and len(retrieved2answer) > 0:
if isinstance(answer2response, (np.ndarray, list)) and len(answer2response) > 0 and \
isinstance(retrieved2response[0], (np.ndarray, list)) and \
len(retrieved2response[0]) > 0 and len(retrieved2answer) > 0:
relevant_retrieved = np.max(retrieved2answer, axis=0, keepdims=True)
relevant_faithful = np.max(relevant_retrieved & retrieved2response, axis=1)
irrelevant_retrieved = ~np.max(retrieved2answer, axis=0, keepdims=True)
Expand Down Expand Up @@ -141,7 +154,9 @@ def evaluate_unfaithfulness(result: RAGResult):
assert result.retrieved2response is not None and result.answer2response is not None
retrieved2response = to_bool(result.retrieved2response)
answer2response = to_bool(result.answer2response)
if len(answer2response) > 0 and len(retrieved2response[0]) > 0:
if isinstance(retrieved2response, (np.ndarray, list)) and len(retrieved2response) > 0 and \
isinstance(retrieved2response[0], (np.ndarray, list)) and \
len(retrieved2response[0]) > 0:
unfaithful = ~np.max(retrieved2response, axis=1)
hallucination = np.mean(unfaithful & ~answer2response)
self_knowledge = np.mean(unfaithful & answer2response)
Expand All @@ -156,7 +171,9 @@ def evaluate_unfaithfulness(result: RAGResult):
def evaluate_faithfulness(result: RAGResult):
assert result.retrieved2response is not None
retrieved2response = to_bool(result.retrieved2response)
if len(retrieved2response) > 0 and len(retrieved2response[0]) > 0:
if isinstance(retrieved2response, (np.ndarray, list)) and len(retrieved2response) > 0 and \
isinstance(retrieved2response[0], (np.ndarray, list)) and \
len(retrieved2response[0]) > 0:
faithful = np.max(retrieved2response, axis=1)
result.metrics[metrics.faithfulness] = np.mean(faithful)
else:
Expand Down