List your company

Post-shift accepted-trajectory risk estimation on Live agent RAG tool benchmark

5.5Post-shift Accepted Trajectory RiskToolChain-CRC
4.312.420.528.6Jun 16, 2026

Evaluation results

MethodLinks
ToolChain-CRCCalibration Strategy=Trajectory calibration2026.065.5
Prompt/retrieval-onlyCalibration Strategy=Prompt/retrieval-only2026.0614.5
Fixed trajectoryCalibration Strategy=Fixed trajectory2026.0622.9
Final-onlyCalibration Strategy=Final-answer-only2026.0635.5