IPAB-AT-2026.01
IPAB (Insilicom PV AI Baseline): Explore drug safety evidence extracted from scientific literature using AI
Scope:
Benchmark Overview
IPAB-AT-2026.01 is the first release of the PV AI Benchmark Initiative, focused on Article Triage (AT) — determining whether a scientific article contains reportable adverse event evidence. This benchmark applies a rigorously evaluated AI model to the entire previous year of PubMed literature across all prescription drugs, generating a large-scale, reproducible reference dataset for pharmacovigilance.
For each drug, the benchmark presents AI-identified safety evidence extracted from peer-reviewed literature. These results are generated independently of sponsor reporting and reflect a consistent, model-driven review of the full literature corpus.
Despite growing adoption of AI in pharmacovigilance, there is no shared standard for evaluating performance at scale. IPAB-AT-2026.01 is designed to establish a transparent, industry-wide baseline for AI-driven literature monitoring.
Benchmark Design & Methodology
This benchmark is built on a manually labeled reference dataset and evaluated using predefined performance metrics before being applied at scale to the full PubMed corpus.
Read the full benchmark designHow to Use These Results
- Compare AI-identified literature coverage against internal monitoring workflows
- Identify potential gaps in historical literature review
- Establish a reference point for evaluating AI tools used in PV operations
- Support internal discussions on AI validation and human-in-the-loop design
- Not a regulatory submission
- Not a replacement for sponsor PV obligations
- Not drug labeling or causality assessment
Drug Dictionary
Browse medications included in the IPAB-AT-2026.01 benchmark and review AI-identified safety evidence from the scientific literature.
Loading drugs...