PIGLeT: Language Grounding Through Neuro-Symbolic Interaction in a 3D World

We propose PIGLeT: a model that learns physical commonsense knowledge through interaction, and then uses this knowledge to ground language. We factorize PIGLeT into a physical dynamics model, and a separate language model. Our dynamics model learns not just what objects are but also what they do: glass cups break when thrown, plastic ones don't. We then use it as the interface to our language model, giving us a unified model of linguistic form and grounded meaning. PIGLeT can read a sentence, simulate neurally what might happen next, and then communicate that result through a literal symbolic representation, or natural language. Experimental results show that our model effectively learns world dynamics, along with how to communicate them. It is able to correctly forecast "what happens next" given an English sentence over 80% of the time, outperforming a 100x larger, text-to-text approach by over 10%. Likewise, its natural language summaries of physical interactions are also judged by humans as more accurate than LM alternatives. We present comprehensive analysis showing room for future work.

CLIPScore: A Reference-free Evaluation Metric for Image Captioning

Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in stark contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker vs reference-based metrics, e.g., news captions that require richer contextual knowledge.

Surface Form Competition: Why the Highest Probability Answer Isn't Always Right

Large language models have shown promising results in zero-shot settings (Brown et al.,2020; Radford et al., 2019). For example, they can perform multiple choice tasks simply by conditioning on a question and selecting the answer with the highest probability. However, ranking by string probability can be problematic due to surface form competition-wherein different surface forms compete for probability mass, even if they represent the same underlying concept, e.g. "computer" and "PC." Since probability mass is finite, this lowers the probability of the correct answer, due to competition from other strings that are valid answers (but not one of the multiple choice options). We introduce Domain Conditional Pointwise Mutual Information, an alternative scoring function that directly compensates for surface form competition by simply reweighing each option according to a term that is proportional to its a priori likelihood within the context of the specific zero-shot task. It achieves consistent gains in zero-shot performance over both calibrated (Zhao et al., 2021) and uncalibrated scoring functions on all GPT-2 and GPT-3 models over a variety of multiple choice datasets.

MultiTalk: A Highly-Branching Dialog Testbed for Diverse Conversations

We study conversational dialog in which there are many possible responses to a given history. We present the MultiTalk Dataset, a corpus of over 320,000 sentences of written conversational dialog that balances a high branching factor (10) with several conversation turns (6) through selective branch continuation. We make multiple contributions to study dialog generation in the highly branching setting. In order to evaluate a diverse set of generations, we propose a simple scoring algorithm, based on bipartite graph matching, to optimally incorporate a set of diverse references. We study multiple language generation tasks at different levels of predictive conversation depth, using textual attributes induced automatically from pretrained classifiers. Our culminating task is a challenging theory of mind problem, a controllable generation task which requires reasoning about the expected reaction of the listener.

Reflective Decoding: Unsupervised Paraphrasing and Abductive Reasoning

Pretrained Language Models (LMs) generate text with remarkable quality, novelty,and coherence. Yet applying LMs to the problems of paraphrasing and infilling currently requires direct supervision, since these tasks break the left-to-right generation setup of pretrained LMs. We present Reflective Decoding, a novel unsupervised approach to apply the capabilities of pretrained LMs to non-sequential tasks. Our approach is general and applicable to two distant tasks - paraphrasing and abductive reasoning. It requires no supervision or parallel corpora, only two pretrained language models: forward and backward. Reflective Decoding operates in two intuitive steps. In the contextualization step, we use LMs to generate many left and right contexts which collectively capture the meaning of the input sentence. Then, in the reflection step we decode in the semantic neighborhood of the input, conditioning on an ensemble of generated contexts with the reverse direction LM. We reflect through the generated contexts, effectively using them as an intermediate meaning representation to generate conditional output. Empirical results demonstrate that Reflective Decoding outperforms strong unsupervised baselines on both paraphrasing and abductive text infilling, significantly narrowing the gap between unsupervised and supervised methods.Reflective Decoding introduces the concept of using generated contexts to represent meaning, opening up new possibilities for unsupervised conditional text generation.

Experience Grounds Language

Successful linguistic communication relies on a shared experience of the world, and it is this shared experience that makes utterances meaningful. Despite the incredible effectiveness of language processing models trained on text alone, today's best systems still make mistakes that arise from a failure to relate language to the physical world it describes and to the social interactions it facilitates. Natural Language Processing is a diverse field, and progress throughout its development has come from new representational theories, modeling techniques, data collection paradigms, and tasks. We posit that the present success of representation learning approaches trained on large text corpora can be deeply enriched from the parallel tradition of research on the contextual and social nature of language. In this article, we consider work on the contextual foundations of language: grounding, embodiment, and social interaction. We describe a brief history and possible progression of how contextual information can factor into our representations, with an eye towards how this integration can move the field forward and where it is currently being pioneered. We believe this framing will serve as a roadmap for truly contextual language understanding.

Evaluating Machines by their Real-World Language Use

There is a fundamental gap between how humans understand and use language -- in open-ended, real-world situations -- and today's NLP benchmarks for language understanding. To narrow this gap, we propose to evaluate machines by their success at real-world language use -- which greatly expands the scope of language tasks that can be measured and studied. We introduce TuringAdvice, a new challenge for language understanding systems. Given a complex situation faced by a real person, a machine must generate helpful advice. We make our challenge concrete by introducing RedditAdvice, a dataset and leaderboard for measuring progress. Though we release a training set with 600k examples, our evaluation is dynamic, continually evolving with the language people use: models must generate helpful advice for recently-written situations. Empirical results show that today's models struggle at our task, even those with billions of parameters. The best model, a finetuned T5, writes advice that is at least as helpful as human-written advice in only 9% of cases. This low performance reveals language understanding errors that are hard to spot outside of a generative setting, showing much room for progress.

BottleSum: Unsupervised and Self-supervised Sentence Summarization using the Information Bottleneck Principle

The principle of the Information Bottleneck (Tishby et al. 1999) is to produce a summary of information X optimized to predict some other relevant information Y. In this paper, we propose a novel approach to unsupervised sentence summarization by mapping the Information Bottleneck principle to a conditional language modelling objective: given a sentence, our approach seeks a compressed sentence that can best predict the next sentence. Our iterative algorithm under the Information Bottleneck objective searches gradually shorter subsequences of the given sentence while maximizing the probability of the next sentence conditioned on the summary. Using only pretrained language models with no direct supervision, our approach can efficiently perform extractive sentence summarization over a large corpus. Building on our unsupervised extractive summarization (BottleSumEx), we then present a new approach to self-supervised abstractive summarization (BottleSumSelf), where a transformer-based language model is trained on the output summaries of our unsupervised method. Empirical results demonstrate that our extractive method outperforms other unsupervised models on multiple automatic metrics. In addition, we find that our self-supervised abstractive model outperforms unsupervised baselines (including our own) by human evaluation along multiple attributes.

Counterfactual Story Reasoning and Generation

Counterfactual reasoning requires predicting how alternative events, contrary to what actually happened, might have resulted in different outcomes. Despite being considered a necessary component of AI-complete systems, few resources have been developed for evaluating counterfactual reasoning in narratives. In this paper, we propose Counterfactual Story Rewriting: given an original story and an intervening counterfactual event, the task is to minimally revise the story to make it compatible with the given counterfactual event. Solving this task will require deep understanding of causal narrative chains and counterfactual invariance, and integration of such story reasoning capabilities into conditional language generation models. We present TimeTravel, a new dataset of 29,849 counterfactual rewritings, each with the original story, a counterfactual event, and human-generated revision of the original story compatible with the counterfactual event. Additionally, we include 80,115 counterfactual "branches" without a rewritten storyline to support future work on semi- or un-supervised approaches to counterfactual story rewriting. Finally, we evaluate the counterfactual rewriting capacities of several competitive baselines based on pretrained language models, and assess whether common overlap and model-based automatic metrics for text generation correlate well with human scores for counterfactual rewriting.

Abductive Commonsense Reasoning

Abductive reasoning is inference to the most plausible explanation. For example, if Jenny finds her house in a mess when she returns from work, and remembers that she left a window open, she can hypothesize that a thief broke into her house and caused the mess, as the most plausible explanation. While abduction has long been considered to be at the core of how people interpret and read between the lines in natural language (Hobbs et al. (1988)), there has been relatively little NLP research in support of abductive natural language inference. We present the first study that investigates the viability of language-based abductive reasoning. We conceptualize a new task of Abductive NLI and introduce a challenge dataset, ART, that consists of over 20k commonsense narrative contexts and 200k explanations, formulated as multiple choice questions for easy automatic evaluation. We establish comprehensive baseline performance on this task based on state-of-the-art NLI and language models, which leads to 68.9% accuracy, well below human performance (91.4%). Our analysis leads to new insights into the types of reasoning that deep pre-trained language models fail to perform -- despite their strong performance on the related but fundamentally different task of entailment NLI -- pointing to interesting avenues for future research.