ResearchBenchmarksTask-oriented Dialogue on MultiWOZ 2.1 (SR, HR)Follow51.24Success Rate (SR)VLK-RL27.590433.730239.8746.0098Apr 25, 2026Evaluation ResultsMethodMethodLinksSuccess Rate (SR)Hit Rate (HR)VLK-RLbackbone=Qwen-14Bbackbone=Qwen-14B2026.0451.243.18VLK-RLbackbone=Qwen-7Bbackbone=Qwen-7B2026.0449.363.07VLK-RLbackbone=GPT-4o-minibackbone=GPT-4o-mini2026.0447.133.04GDP-Zero2026.0441.22.61CAPID2026.0440.562.63TransferTOD2026.0439.822.58GALAXY2026.0438.752.46ACGOS2026.0434.12.35PPO2026.0428.52.21