FaithDial: A Faithful Benchmark for Information-Seeking Dialogue
This addresses the issue of unreliable responses in dialogue systems for users seeking accurate information, though it is incremental as it builds on existing benchmarks.
The authors tackled the problem of hallucination in information-seeking dialogue systems by creating FaithDial, a benchmark derived from Wizard of Wikipedia, which reduces unsupported utterances while maintaining engaging conversations. They demonstrated its utility by training models that achieve a 12.8 F1 score improvement on hallucination detection and show better performance in generation metrics and human evaluations.
The goal of information-seeking dialogue is to respond to seeker queries with natural language utterances that are grounded on knowledge sources. However, dialogue systems often produce unsupported utterances, a phenomenon known as hallucination. To mitigate this behavior, we adopt a data-centric solution and create FaithDial, a new benchmark for hallucination-free dialogues, by editing hallucinated responses in the Wizard of Wikipedia (WoW) benchmark. We observe that FaithDial is more faithful than WoW while also maintaining engaging conversations. We show that FaithDial can serve as training signal for: i) a hallucination critic, which discriminates whether an utterance is faithful or not, and boosts the performance by 12.8 F1 score on the BEGIN benchmark compared to existing datasets for dialogue coherence; ii) high-quality dialogue generation. We benchmark a series of state-of-the-art models and propose an auxiliary contrastive objective that achieves the highest level of faithfulness and abstractiveness based on several automated metrics. Further, we find that the benefits of FaithDial generalize to zero-shot transfer on other datasets, such as CMU-Dog and TopicalChat. Finally, human evaluation reveals that responses generated by models trained on FaithDial are perceived as more interpretable, cooperative, and engaging.