Holistic Evaluation of Language Models

Language models (LMs) are becoming the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of language models. First, we taxonomize the vast space of potential scenarios (i.e. use cases) and metrics (i.e. desiderata) that are of interest for LMs. Then we select a broad subset based on coverage and feasibility, noting what's missing or underrepresented (e.g. question answering for neglected English dialects, metrics for trustworthiness). Second, we adopt a multi-metric approach: We measure 7 metrics (accuracy, calibration, robustness, fairness, bias, toxicity, and efficiency) for each of 16 core scenarios when possible (87.5% of the time). This ensures metrics beyond accuracy don't fall to the wayside, and that trade-offs are clearly exposed. We also perform 7 targeted evaluations, based on 26 targeted scenarios, to analyze specific aspects (e.g. reasoning, disinformation). Third, we conduct a large-scale evaluation of 30 prominent language models (spanning open, limited-access, and closed models) on all 42 scenarios, 21 of which were not previously used in mainstream LM evaluation. Prior to HELM, models on average were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: now all 30 models have been densely benchmarked on the same core scenarios and metrics under standardized conditions. Our evaluation surfaces 25 top-level findings. For full transparency, we release all raw model prompts and completions publicly for further analysis, as well as a general modular toolkit. We intend for HELM to be a living benchmark for the community, continuously updated with new scenarios, metrics, and models.

When Do You Need Billions of Words of Pretraining Data?

NLP is currently dominated by general-purpose pretrained language models like RoBERTa, which achieve strong performance on NLU tasks through pretraining on billions of words. But what exact knowledge or skills do Transformer LMs learn from large-scale pretraining that they cannot learn from less data? We adopt four probing methods---classifier probing, information-theoretic probing, unsupervised relative acceptability judgment, and fine-tuning on NLU tasks---and draw learning curves that track the growth of these different measures of linguistic ability with respect to pretraining data volume using the MiniBERTas, a group of RoBERTa models pretrained on 1M, 10M, 100M and 1B words. We find that LMs require only about 10M or 100M words to learn representations that reliably encode most syntactic and semantic features we test. A much larger quantity of data is needed in order to acquire enough commonsense knowledge and other skills required to master typical downstream NLU tasks. The results suggest that, while the ability to encode linguistic features is almost certainly necessary for language understanding, it is likely that other forms of knowledge are the major drivers of recent improvements in language understanding among large pretrained models.

Learning Which Features Matter: RoBERTa Acquires a Preference for Linguistic Generalizations (Eventually)

One reason pretraining on self-supervised linguistic tasks is effective is that it teaches models features that are helpful for language understanding. However, we want pretrained models to learn not only to represent linguistic features, but also to use those features preferentially during fine-turning. With this goal in mind, we introduce a new English-language diagnostic set called MSGS (the Mixed Signals Generalization Set), which consists of 20 ambiguous binary classification tasks that we use to test whether a pretrained model prefers linguistic or surface generalizations during fine-tuning. We pretrain RoBERTa models from scratch on quantities of data ranging from 1M to 1B words and compare their performance on MSGS to the publicly available RoBERTa-base. We find that models can learn to represent linguistic features with little pretraining data, but require far more data to learn to prefer linguistic generalizations over surface ones. Eventually, with about 30B words of pretraining data, RoBERTa-base does demonstrate a linguistic bias with some regularity. We conclude that while self-supervised pretraining is an effective way to learn helpful inductive biases, there is likely room to improve the rate at which models learn which features matter.

Latent Tree Learning with Ordered Neurons: What Parses Does It Produce?

Recent latent tree learning models can learn constituency parsing without any exposure to human-annotated tree structures. One such model is ON-LSTM (Shen et al., 2019), which is trained on language modelling and has near-state-of-the-art performance on unsupervised parsing. In order to better understand the performance and consistency of the model as well as how the parses it generates are different from gold-standard PTB parses, we replicate the model with different restarts and examine their parses. We find that (1) the model has reasonably consistent parsing behaviors across different restarts, (2) the model struggles with the internal structures of complex noun phrases, (3) the model has a tendency to overestimate the height of the split points right before verbs. We speculate that both problems could potentially be solved by adopting a different training task other than unidirectional language modelling.