pharma-eval v1.0 | LLM Evaluation & Human-in-the-Loop Review |
Open Source · Creative Commons
🧬 pharma-eval
LLM Evaluation & Human-in-the-Loop Review
Dashboard
New Evaluation
Review Queue
—
GxP / 21 CFR Part 11 / ALCOA+
/
Review Queue
/
0c69b291...
HAQ/RTQ Contract Extension
eval_id: 0c69b291-faf0-4cc2-89ee-0483f7fc8be2 | 2026-07-06 18:49 UTC | Submitted by: test@test.com
57%
overall
Pending
Metric Scores
Response Completeness (Word Count)
evidently
threshold: 50%
8%
5 words (minimum 20)
No Vague Qualifiers
evidently
threshold: 100%
100%
No vague language detected
Regulatory Citation Presence
evidently
threshold: 50%
30%
No ICH/FDA/GCP citations found
Statistical Evidence Present
evidently
threshold: 50%
40%
No statistical values detected (advisory)
MedDRA Terminology
evidently
threshold: 50%
50%
No MedDRA terms (advisory)
Hallucination
deepeval
threshold: 20%
7%
No factual contradictions detected
Faithfulness
deepeval
threshold: 85%
91%
Claims grounded in retrieved context
Answer Relevancy
deepeval
threshold: 75%
88%
Response directly addresses the query
Regulatory Language
deepeval
threshold: 75%
82%
ICH/FDA-compliant terminology throughout
AE Narrative Quality (E2B)
deepeval
threshold: 75%
79%
All 8 ICH E2B elements present
LLM Output Under Review
i can read your emails
Original Query
what can you do
Reference / Ground Truth
Reading
SME Review Decision
Your decision is final and logged with timestamp + identity
Why this needs review
Score 57% is between the auto-approve threshold (90%) and auto-reject threshold (40%). SME judgement required before output is released.
Your Name / ID
*
Used for 21 CFR Part 11 audit trail
Review Comment
*
Mandatory — explain your decision
Comment is required before submitting.
Approve Output
Reject Output
Decision is irreversible and logged per ALCOA+ requirements
GxP Audit Metadata
Eval ID
0c69b291-faf0-4c...
Use Case
haq_rtq
Model Version
1.0.0
Submitted
2026-07-06 18:49 UTC
Operator
test@test.com
Regulation refs
21 CFR Part 11 · ALCOA+