ResearchBenchmarksDialogue Response Generation on PhotoChat (test)Follow0.68KappaDivter0.6280.64150.6550.6685Oct 16, 2021Evaluation ResultsMethodMethodLinksKappaWLTContext CoherenceText FluencyImage QualityBackground ConsistencyDivtervariant=multimodal, co...variant=multimodal, comparison baseline=DialoGPT2021.100.6853.527.419.1————S2S-TF2021.100.67———0.420.580.2520SCAN2021.100.65—————1.9596Divtervariant=pure text, com...variant=pure text, comparison baseline=DialoGPT2021.100.6434.435.729.9————Divter2021.100.63———1.591.831.63161