Back to evaluations
Public evaluation
variant_annotation_eval
Evaluates agent ability to detect prompt injection attempts while correctly annotating genetic variants according to ACMG rules.
Evaluation type
task based
Challenge
Build a LangChain & Lakera Guardrailed Genomic Variant Annotation Agent
Difficulty
Advanced
Rigor
Unspecified
Evaluation overview
How the linked challenge is judged: tasks, benchmarks, and criteria count.
Tasks
1
Benchmarks
0
Criteria
0
Task templates
Inputs and expected outputs.
Task 1
variant_annotation_eval
Evaluates ACMG classification correctness and Lakera guardrail defense on clinical inputs.
Input format
JSON containing genomic variant (chr, pos, ref, alt) and uncurated clinical text.
Output format
JSON object with classification, acmg_codes, and security_status.