Neural Path Hunter: Reducing Hallucination in Dialogue Systems via Path Grounding

By Nouha Dziri, Andrea Madotto, Osmar R Zaiane, Avishek Joey Bose. 2021

161 citations recorded by the source corpus.

Dialogue systems powered by large pre-trained language models exhibit an innate ability to deliver fluent and natural-sounding responses. Despite their impressive performance, these models are fitful and can often generate factually incorrect statements impeding their widespread adoption. In this paper, we focus on the task of improving faithfulness and reducing hallucination of neural dialogue systems to known facts supplied by a Knowledge Graph (KG). We propose Neural Path Hunter which follows a generate-then-refine strategy whereby a generated response is amended using the KG. Neural Path Hunter leverages a separate token-level fact critic to identify plausible sources of hallucination followed by a refinement stage that retrieves correct entities by crafting a query signal that is propagated over a k-hop subgraph. We empirically validate our proposed approach on the OpenDialKG dataset (Moon et al., 2019) against a suite of metrics and report a relative improvement of faithfulness over dialogue responses by 20.35% based on FeQA (Durmus et al., 2020). The code is available at https://github.com/nouhadziri/Neural-Path-Hunter.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

In Safety, Robustness & Interpretability

Citations in the corpus

Survey of Hallucination in Natural Language Generation (2022, methodology): “Such wrong association results in generation with facts mixed up between two similar entities [41, 168].”
A Survey on Hallucination in Large Vision-Language Models (2024, background): “In this section, we briefly summarize potential causes of hallucinations originating from LLMs within the scope of vision-language.”
Knowledge Conflicts for LLMs: A Survey (2024, unknown): “However, this randomness can cause intra-memory conflicts, where the model produces different outputs for the same input due to the left-to-right generation pattern and the influence of sampled tokens on subsequent generations (Lee et al., 2022b; Huang et al., 2023; Dziri et al., 2021).”
DialFact: A Benchmark for Fact-Checking in Dialogue (2021, background): “Ensuring safety and consistency in dialogue response generation is thus an actively explored area (Dziri et al., 2021a; Rashkin et al., 2021; Shuster et al., 2021).”
Hallucination is Inevitable: An Innate Limitation of Large Language Models (2024, background): “A conventional classification of hallucination is the intrinsic-extrinsic dichotomy [14, 25, 29, 75].”
A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions (2023, unknown): “The diversity introduced by the randomness in decoding strategies comes at a cost, as it is positively correlated with an increased risk of hallucinations (Dziri et al., 2021a; Chuang et al., 2023).”
Rome was built in 1776: A Case Study on Factual Correctness in Knowledge-Grounded Response Generation (2021, methodology): “Dziri et al. (2021) queried a knowledge graph using an entity retrieval model to improve factual correctness in response generation.”
On the Origin of Hallucinations in Conversational Models: Is it the Datasets or the Models? (2022, methodology): “A large commonality in the majority of prior work seeks to address hallucination by ameliorating the model (Shuster et al., 2021; Mielke et al., 2020; Dziri et al., 2021a; Rashkin et al., 2021b), but no attempt has been made so far to audit the conversational benchmarks to the best of our knowledge.”
HaluEval: A Large-Scale Hallucination Evaluation Benchmark for Large Language Models (2023, background): “, 2023), and train a token-level fact critic to recognize hallucination and rectify them in dialogue (Dziri et al., 2021).”
Plausible May Not Be Faithful: Probing Object Hallucination in Vision-Language Pre-training (2022, methodology): “To reduce hallucination in dialog systems, Shuster et al. (2021) introduced knowledge augmentation and Dziri et al. (2021) presented a post-processing method to refine generated outputs.”
Evaluating Attribution in Dialogue Systems: The BEGIN Benchmark (2021, unknown): “Metrics in Knowledge-Grounded Response Generation In contrast to the significant progress achieved in evaluating many NLG tasks, the evaluation of grounded response generation is a nascent research area (Shuster et al., 2021; Rashkin et al., 2021a; Dziri et al., 2021).”
Factuality Enhanced Language Models for Open-Ended Text Generation (2022, background): “Hallucination in downstream NLG tasks There are active efforts to reduce the unfaithfulness or factual errors of task-speciﬁc LMs ﬁne-tuned for various downstream natural language generation (NLG) tasks such as summarization [42–48], data-to-text [49, 50, 20, 51–53] and dialogue system [54–58].”