Skip to content

README.md

Latest commit

 

History

History
24 lines (19 loc) · 1.17 KB

README.md

File metadata and controls

24 lines (19 loc) · 1.17 KB
Raw

JavaScript control audits

These records exercise independent Node.js references and deliberately faulty variants through the existing task graders. They use seed 17 in Docker.

Task Reference Detected faults Evidence
Durable KV 15/15 cases 8/8 Report, JSON, events
Simulated support 4/4 cases 7/7 Report, JSON, events

Reproduce after installing EvalArc 0.6 and pulling node:22-slim:

evalarc audit --language javascript --image node:22-slim \
  --seeds 17 --output runs/js-coding-audit
evalarc audit --task support-routing --language javascript --image node:22-slim \
  --seeds 17 --output runs/js-support-audit

The validation record identifies the image used. Each JSON preserves the resolved image ID and the full per-control evaluation. The archived output is evidence of these executions, not a model benchmark or proof of general grader robustness. The original Python examples retain their historical provenance.