CAIN-42 CAIN StudioProduct #13 — Now Available
CAIN Eval
Agent evaluation platform. Trajectory grading and contamination detection for production AI systems.
Evaluation Statistics
-
Total Gradings
-
Avg Score
-
Pass
-
Conditional
-
Fail
Evaluation Tools
Grade a successful agent trajectory for path quality. Checks: redundant calls, wasted cost, risky-then-corrected patterns.
Scan for benchmark contamination using membership-inference style probing. Compares benchmark vs control set overlap.
TrajectoryGate
Deterministic grading of successful agent runs. Scores path quality: redundant calls, wasted cost, risky corrections.
Contamination Scanner
Benchmark contamination detection using membership-inference style probing. Compares benchmark vs control overlap.
Eval Reports
Comprehensive evaluation reports with recommendations for agent optimization and improvement.