Controlled Hallucinations: Learning to Generate Faithfully from Noisy Data

By Katja Filippova. 2020

145 citations recorded by the source corpus.

Neural text generation (data- or text-to-text) demonstrates remarkable performance when training data is abundant which for many applications is not the case. To collect a large corpus of parallel data, heuristic rules are often used but they inevitably let noise into the data, such as phrases in the output which cannot be explained by the input. Consequently, models pick up on the noise and may hallucinate–generate fluent but unsupported text. Our contribution is a simple but powerful technique to treat such hallucinations as a controllable aspect of the generated text, without dismissing any input and without modifying the model architecture. On the WikiBio corpus (Lebret et al., 2016), a particularly noisy dataset, we demonstrate the efficacy of the technique both in an automatic and in a human evaluation.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing, Generative Models

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 Generative Models

Citations in the corpus

Survey of Hallucination in Natural Language Generation (2022, background): “Controllable generation techniques such as controlled re-sampling [152], control codes that can be provided manually [50, 152, 210], or predicted automatically [210] are leveraged to improve faithfulness.”
Hallucinated but Factual! Inspecting the Factuality of Hallucinations in Abstractive Summarization (2021, methodology): “(4) The LM-based method is proposed by Filippova (2020). The LM-based method uses LM and CLM to compute the token’s prior and posterior probability. In Filippova (2020)’s work, they compare the value of pprior and ppos. If the generated token does not match the reference and pprior is greater than p”
Controlling hallucinations at word level in data-to-text generation (2021, methodology): “(2) Unlike previous CTG approaches [32, 60, 11, 12] which propose instance-level control factors, the control of the hallucination factor is performed at the word-level to enable ﬁner-grained signal to be sent to the model.”
Improving Faithfulness in Abstractive Summarization with Contrast Candidate Generation and Selection (2021, background): “light the existence of hallucinations in training data, truncating potentially unfaithful gold summaries during training is an effective strategy (Kang and Hashimoto, 2020; Filippova, 2020).”
Increasing Faithfulness in Knowledge-Grounded Dialogue with Controllable Features (2021, methodology): “Similar to our approach, Filippova (2020) also uses control codes to reduce hallucinations but focused instead on data-to-text generation tasks.”
Detecting hallucinations in large language models using semantic entropy (2024, unknown): “Hallucinations are often defined as LLMs generating “content that is nonsensical or unfaithful to the provided source content” 9–11 but they have come to include a vast array of failures of faithfulness and factuality.”
TaTa: A Multilingual Table-to-Text Dataset for African Languages (2022, methodology): “Second, we use a tagging approach suggested by Filippova (2020) where we append a “0” to the input for examples without a reference and learn to predict an empty string.”
Sketch and Refine: Towards Faithful and Informative Table-to-Text Generation (2021, background): “Filippova (2020) introduces a “hallucination knob” to reduce the amount of hallucinations in the generated text.”
A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions (2023, unknown): “Within the realm of NLP, hallucination is typically referred to as a phenomenon in which the generated content appears nonsensical or unfaithful to the provided source content (Filippova, 2020; Maynez et al., 2020).”
CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing (2023, background): “, factual consistent with the provided source content (Filippova, 2020; Zhou et al., 2021).”