ResearchBenchmarksResponse Generation on Dialogue datasetResponse Generation on Dialogue datasetFollow3.67CoherenceGround Truth2.81723.03863.263.4814Apr 8, 2026Evaluation resultsMethodMethodLinksCoherenceFluencyInformativenessHelpfulnessOverall ScoreGround Truth2026.043.674.423.814.324.06Qwen3-8BFine-tuning=DRCRFine-tuning=DRCR2026.043.414.363.874.123.94SS-MPC2026.043.184.123.533.873.68Qwen3-8BFine-tuning=SFTFine-tuning=SFT2026.043.124.033.513.833.62RL-TRC2026.043.013.893.43.623.48MADNet2026.042.853.643.193.373.26