Most existing dialogue systems fail to respond properly to potentially unsafe user utterances by either ignoring or passively agreeing with them.
13 PAPERS • 1 BENCHMARK
SODA is a high-quality social dialogue dataset. In contrast to most existing crowdsourced, small-scale dialogue corpora, Soda distills 1.5M socially-grounded dialogues from a pre-trained language model (InstructGPT; Ouyang et al., ). Dialogues are distilled by contextualizing social commonsense knowledge from a knowledge graph (Atomic10x).
11 PAPERS • NO BENCHMARKS YET