Despite their ability to capture large amount of knowledge during pretraining, large-scale language models often benefit from incorporating external knowledge bases, especially on commonsense reasoning tasks. This motivates us to explore how we can best leverage knowledge elicited from language models themselves. We propose generating knowledge statements directly from a language model with a generic prompt format, then selecting the knowledge which maximizes prediction probability. Despite its simplicity, this approach improves performance of both off-the-shelf and finetuned language models on four commonsense reasoning tasks, improving the state-of-the-art on numerical commonsense (NumerSense), general commonsense (CommonsenseQA 2.0), and scientific commonsense (QASC) benchmarks. Notably, we find that a model's predictions can improve when using its own generated knowledge, demonstrating the importance of symbolic knowledge representation in neural reasoning processes.
How can model designers turn task instructions into effective prompts for language models? Backed by extensive empirical analysis on GPT3, we observe important features for successful instructional prompts, and propose several reframing techniques for model designers to create such prompts. For example, a complex task can be decomposed into multiple simpler tasks. We experiment over 12 NLP tasks across 6 diverse categories (question generation, classification, etc.). Our results show that reframing improves few-shot learning performance by 14\% while reducing sample complexity over existing few-shot baselines. The performance gains are particularly important on large language models, such as GPT3 where tuning models or prompts on large datasets is not feasible. Furthermore, we observe that such gains are not limited to GPT3; the reframed tasks remain superior over raw instructions across different model architectures, underscoring the cross-model generality of these guidelines. We hope these empirical-driven techniques will pave way for more effective ways to prompt LMs in future.
Identifying relevant knowledge to be used in conversational systems that are grounded in long documents is critical to effective response generation. We introduce a knowledge identification model that leverages the document structure to provide dialogue-contextualized passage encodings and better locate knowledge relevant to the conversation. An auxiliary loss captures the history of dialogue-document connections. We demonstrate the effectiveness of our model on two document-grounded conversational datasets and provide analyses showing generalization to unseen documents and long dialogue contexts.
Large pre-trained models such as CLIP offer consistent accuracy across a range of data distributions when performing zero-shot inference (i.e., without fine-tuning on a specific dataset). Although existing fine-tuning approaches substantially improve accuracy in-distribution, they also reduce out-of-distribution robustness. We address this tension by introducing a simple and effective method for improving robustness: ensembling the weights of the zero-shot and fine-tuned models. Compared to standard fine-tuning, the resulting weight-space ensembles provide large accuracy improvements out-of-distribution, while matching or improving in-distribution accuracy. On ImageNet and five derived distribution shifts, weight-space ensembles improve out-of-distribution accuracy by 2 to 10 percentage points while increasing in-distribution accuracy by nearly 1 percentage point relative to standard fine-tuning. These improvements come at no additional computational cost during fine-tuning or inference.
We introduce a noisy channel approach for language model prompting in few-shot text classification. Instead of computing the likelihood of the label given the input (referred as direct models), channel models compute the conditional probability of the input given the label, and are thereby required to explain every word in the input. We use channel models for recently proposed few-shot learning methods with no or very limited updates to the language model parameters, via either in-context demonstration or prompt tuning. Our experiments show that, for both methods, channel models significantly outperform their direct counterparts, which we attribute to their stability, i.e., lower variance and higher worst-case accuracy. We also present extensive ablations that provide recommendations for when to use channel prompt tuning instead of other competitive models (e.g., direct head tuning): channel prompt tuning is preferred when the number of training examples is small, labels in the training data are imbalanced, or generalization to unseen labels is required.
We present CORA, a Cross-lingual Open-Retrieval Answer Generation model that can answer questions across many languages even when language-specific annotated data or knowledge sources are unavailable. We introduce a new dense passage retrieval algorithm that is trained to retrieve documents across languages for a question. Combined with a multilingual autoregressive generation model, CORA answers directly in the target language without any translation or in-language retrieval modules as used in prior work. We propose an iterative training method that automatically extends annotated data available only in high-resource languages to low-resource ones. Our results show that CORA substantially outperforms the previous state of the art on multilingual open question answering benchmarks across 26 languages, 9 of which are unseen during training. Our analyses show the significance of cross-lingual retrieval and generation in many languages, particularly under low-resource settings.
Despite significant interest in developing general purpose fact checking models, it is challenging to construct a large-scale fact verification dataset with realistic claims that would occur in the real world. Existing claims are either authored by crowdworkers, thereby introducing subtle biases that are difficult to control for, or manually verified by professional fact checkers, causing them to be expensive and limited in scale. In this paper, we construct a challenging, realistic, and large-scale fact verification dataset called FaVIQ, using information-seeking questions posed by real users who do not know how to answer. The ambiguity in information-seeking questions enables automatically constructing true and false claims that reflect confusions arisen from users (e.g., the year of the movie being filmed vs. being released). Our claims are verified to be natural, contain little lexical bias, and require a complete understanding of the evidence for verification. Our experiments show that the state-of-the-art models are far from solving our new task. Moreover, training on our data helps in professional fact-checking, outperforming models trained on the most widely used dataset FEVER or in-domain data by up to 17% absolute. Altogether, our data will serve as a challenging benchmark for natural language understanding and support future progress in professional fact checking.
Biomedical knowledge graphs (KGs) hold rich information on entities such as diseases, drugs, and genes. Predicting missing links in these graphs can boost many important applications, such as drug design and repurposing. Recent work has shown that general-domain language models (LMs) can serve as "soft" KGs, and that they can be fine-tuned for the task of KG completion. In this work, we study scientific LMs for KG completion, exploring whether we can tap into their latent knowledge to enhance biomedical link prediction. We evaluate several domain-specific LMs, fine-tuning them on datasets centered on drugs and diseases that we represent as KGs and enrich with textual entity descriptions. We integrate the LM-based models with KG embedding models, using a router method that learns to assign each input example to either type of model and provides a substantial boost in performance. Finally, we demonstrate the advantage of LM models in the inductive setting with novel scientific entities. Our datasets and code are made publicly available.
Many commonsense reasoning NLP tasks involve choosing between one or more possible answers to a question or prompt based on knowledge that is often implicit. Large pretrained language models (PLMs) can achieve near-human performance on such tasks, while providing little human-interpretable evidence of the underlying reasoning they use. In this work, we show how to use these same models to generate such evidence: inspired by the contrastive nature of human explanations, we use PLMs to complete explanation prompts which contrast alternatives according to the key attribute(s) required to justify the correct answer (for example, peanuts are usually salty while raisins are sweet). Conditioning model decisions on these explanations improves performance on two commonsense reasoning benchmarks, as compared to previous non-contrastive alternatives. These explanations are also judged by humans to be more relevant for solving the task, and facilitate a novel method to evaluate explanation faithfulfness.