ArgSciChat: A Dataset for Argumentative Dialogues on Scientific Papers

Benchmark Model Rank Results
fact-selection-on-argscichatTF-IDF#1Fact-F1: 16.22
fact-selection-on-argscichatS-BERT#2Fact-F1: 13.65
fact-selection-on-argscichatLED(Q,P)#3Fact-F1: 10.58
fact-selection-on-argscichatLED(Q,P,H)#4Fact-F1: 8.50
response-generation-on-argscichatLED(Q,F)#1Message-F1: 19.54BScore: 86.64Mover: 8.53
response-generation-on-argscichatLED(Q,P,H)#2Message-F1: 16.14BScore: 86.00Mover: 4.54
response-generation-on-argscichatLED(Q,P)#3Message-F1: 14.25BScore: 85.85Mover: 2.25