How Context Affects Language Models' Factual Predictions

By Fabio Petroni, Patrick Lewis, Aleksandra Piktus, Tim Rocktaschel, Yuxiang Wu, Alexander H. Miller, Sebastian Riedel. 2020

271 citations recorded by the source corpus.

When pre-trained on large unsupervised textual corpora, language models are able to store and retrieve factual knowledge to some extent, making it possible to use them directly for zero-shot cloze-style question answering. However, storing factual knowledge in a fixed number of weights of a language model clearly has limitations. Previous approaches have successfully provided access to information outside the model weights using supervised architectures that combine an information retrieval system with a machine reading component. In this paper, we go a step further and integrate information from a retrieval system with a pre-trained language model in a purely unsupervised way. We report that augmenting pre-trained language models in this way dramatically improves performance and that the resulting system, despite being unsupervised, is competitive with a supervised machine reading baseline. Furthermore, processing query and context with different segment tokens allows BERT to utilize its Next Sentence Prediction pre-trained classifier to determine whether the context is relevant or not, substantially improving BERT's zero-shot cloze-style question-answering performance and making its predictions robust to noisy contexts.

Subfields: Large Language Models

How researchers see this paper

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

In Large Language Models

Citations in the corpus

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020, background): “Such tasks include open-domain question answering [5, 25], fact checking [50], fact completion [42], long-form question answering [12], Wikipedia article generation [32], dialogue [36, 59, 9, 13], translation [16], and language modeling [17, 23].”
Few-shot Learning with Retrieval Augmented Language Models (2022, methodology): “First, we use the KILT evaluation suite (Petroni et al., 2020), containing 11 datasets corresponding to 5 tasks: fact checking, question answering, dialog generation, entity linking and slot-filling.”
Autoregressive Search Engines: Generating Substrings as Document Identifiers (2022, background): “Some works have demonstrated that knowledge stored in the parameters of these models can be retrieved to some extend by directly generating evidence given a query (Petroni et al., 2019, 2020; Roberts et al., 2020).”
Autoregressive Entity Retrieval (2020, methodology): “The idea of using a generative model for entity disambiguation was proposed in Petroni et al. (2020b) as they trained both BART and T5 in a seq2seq fashion on all KILT tasks (including ED).”
Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing (2021, methodology): “Prompt augmentation is closely related to retrieval-based methods that provide more textual context to the model to improve performance [37], a method that has also been shown to be effective in prompt-based learning [99].”
KILT: a Benchmark for Knowledge Intensive Language Tasks (2020, methodology): “For our BART+DPR baseline, we follow Petroni et al. (2020) to retrieve and prepend the top-3 passages from DPR for each input sample and use context-enhanced training data to ﬁne-tune a BART model.”
Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity (2021, background): “…et al., 2019; Peters et al., 2018; Raffel et al., 2020; Liu et al., 2019; Yang et al., 2019; Radford et al., 2019) have shown remarkable performance when conditioned with an appropriate textual context (Petroni et al., 2019, 2020; Jiang et al., 2020; Shin et al., 2020; Davison et al., 2019).”
Locating and Editing Factual Associations in GPT (2022, background): “A common strategy is to deﬁne a ﬁll-in-the-blank prompt, and let a masked LM complete it (Petroni et al., 2019, 2020).”
GPT Understands, Too (2021, background): “LAMA (Petroni et al., 2019, 2020) propose the LAMA task that leverages cloze tests to predict the fact triples of knowledge bases to examine language model’s ability of memorizing facts with answers in the single-token format.”
Mass-Editing Memory in a Transformer (2022, background): “, 2022) are capable of recalling an impressive array of common facts such as “Tim Cook is the CEO of Apple” or “Polaris is in the constellation Ursa Minor” (Petroni et al., 2020; Brown et al., 2020), even very large models are known to lack more specialized knowledge, and they may recall obsolete in”
CrossFit: A Few-shot Learning Challenge for Cross-task Generalization in NLP (2021, background): “…lama-conceptnet qa/closed-book qa Petroni et al. 2019, 2020 lama-google_re qa/closed-book qa Petroni et al. 2019, 2020 lama-squad qa/closed-book qa Petroni et al. 2019, 2020 lama-trex qa/closed-book qa Petroni et al. 2019, 2020 liar cls/fact checking Wang 2017 limit other Manotas et al. 2020…”
X-FACTR: Multilingual Factual Knowledge Retrieval from Pretrained Language Models (2020, background): “Petroni et al. (2020) further shows that LMs can generate even more factual knowledge when augmented with retrieved sentences.”
Robust Retrieval Augmented Generation for Zero-shot Slot Filling (2021, methodology): “Recently, pre-trained language models have been used for slot filling (Petroni et al., 2020a), opening a new research direction that might provide an effective solution to the aforementioned problems.”
BERT-kNN: Adding a kNN Search Component to Pretrained Language Models for Better QA (2020, methodology): “Following Petroni et al. (2019) we report mean precision at rank r (P@r).”
Learning Knowledge Bases with Parameters for Task-Oriented Dialogue Systems (2020, background): “Moreover, several studies tried to extract (Petroni et al., 2019; Kassner and Schütze, 2019; Petroni et al., 2020) or use (Roberts et al.”