Joint Retrieval and Generation Training for Grounded Text Generation

Recent advances in large-scale pre-training such as GPT-3 allow seemingly high quality text to be generated from a given prompt. However, such generation systems often suffer from problems of hallucinated facts, and are not inherently designed to incorporate useful external information. Grounded generation models appear to offer remedies, but their training typically relies on rarely-available parallel data where corresponding information-relevant documents are provided for context. We propose a framework that alleviates this data constraint by jointly training a grounded generator and document retriever on the language model signal. The model learns to reward retrieval of the documents with the highest utility in generation, and attentively combines them using a Mixture-of-Experts (MoE) ensemble to generate follow-on text. We demonstrate that both generator and retriever can take advantage of this joint training and work synergistically to produce more informative and relevant text in both prose and dialogue generation.

A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation

Large pretrained generative models like GPT-3 often suffer from hallucinating non-existent or incorrect content, which undermines their potential merits in real applications. Existing work usually attempts to detect these hallucinations based on a corresponding oracle reference at a sentence or document level. However ground-truth references may not be readily available for many free-form text generation applications, and sentence- or document-level detection may fail to provide the fine-grained signals that would prevent fallacious content in real time. As a first step to addressing these issues, we propose a novel token-level, reference-free hallucination detection task and an associated annotated dataset named HaDes (HAllucination DEtection dataSet). To create this dataset, we first perturb a large number of text segments extracted from English language Wikipedia, and then verify these with crowd-sourced annotations. To mitigate label imbalance during annotation, we utilize an iterative model-in-loop strategy. We conduct comprehensive data analyses and create multiple baseline models.

An Adversarially-Learned Turing Test for Dialog Generation Models

The design of better automated dialogue evaluation metrics offers the potential of accelerate evaluation research on conversational AI. However, existing trainable dialogue evaluation models are generally restricted to classifiers trained in a purely supervised manner, which suffer a significant risk from adversarial attacking (e.g., a nonsensical response that enjoys a high classification score). To alleviate this risk, we propose an adversarial training approach to learn a robust model, ATT (Adversarial Turing Test), that discriminates machine-generated responses from human-written replies. In contrast to previous perturbation-based methods, our discriminator is trained by iteratively generating unrestricted and diverse adversarial examples using reinforcement learning. The key benefit of this unrestricted adversarial training approach is allowing the discriminator to improve robustness in an iterative attack-defense game. Our discriminator shows high accuracy on strong attackers including DialoGPT and GPT-3.

InverseForm: A Loss Function for Structured Boundary-Aware Segmentation

We present a novel boundary-aware loss term for semantic segmentation using an inverse-transformation network, which efficiently learns the degree of parametric transformations between estimated and target boundaries. This plug-in loss term complements the cross-entropy loss in capturing boundary transformations and allows consistent and significant performance improvement on segmentation backbone models without increasing their size and computational complexity. We analyze the quantitative and qualitative effects of our loss function on three indoor and outdoor segmentation benchmarks, including Cityscapes, NYU-Depth-v2, and PASCAL, integrating it into the training phase of several backbone networks in both single-task and multi-task settings. Our extensive experiments show that the proposed method consistently outperforms baselines, and even sets the new state-of-the-art on two datasets.

Finetuning Pretrained Transformers into RNNs

Transformers have outperformed recurrent neural networks (RNNs) in natural language generation. This comes with a significant computational overhead, as the attention mechanism scales with a quadratic complexity in sequence length. Efficient transformer variants have received increasing interest from recent works. Among them, a linear-complexity recurrent variant has proven well suited for autoregressive generation. It approximates the softmax attention with randomized or heuristic feature maps, but can be difficult to train or yield suboptimal accuracy. This work aims to convert a pretrained transformer into its efficient recurrent counterpart, improving the efficiency while retaining the accuracy. Specifically, we propose a swap-then-finetune procedure: in an off-the-shelf pretrained transformer, we replace the softmax attention with its linear-complexity recurrent alternative and then finetune. With a learned feature map, our approach provides an improved tradeoff between efficiency and accuracy over the standard transformer and other recurrent variants. We also show that the finetuning process needs lower training cost than training these recurrent variants from scratch. As many recent models for natural language tasks are increasingly dependent on large-scale pretrained transformers, this work presents a viable approach to improving inference efficiency without repeating the expensive pretraining process.

Data Augmentation for Abstractive Query-Focused Multi-Document Summarization

The progress in Query-focused Multi-Document Summarization (QMDS) has been limited by the lack of sufficient largescale high-quality training datasets. We present two QMDS training datasets, which we construct using two data augmentation methods: (1) transferring the commonly used single-document CNN/Daily Mail summarization dataset to create the QMDSCNN dataset, and (2) mining search-query logs to create the QMDSIR dataset. These two datasets have complementary properties, i.e., QMDSCNN has real summaries but queries are simulated, while QMDSIR has real queries but simulated summaries. To cover both these real summary and query aspects, we build abstractive end-to-end neural network models on the combined datasets that yield new state-of-the-art transfer results on DUC datasets. We also introduce new hierarchical encoders that enable a more efficient encoding of the query together with multiple documents. Empirical results demonstrate that our data augmentation and encoding methods outperform baseline models on automatic metrics, as well as on human evaluations along multiple attributes.

What Makes Good In-Context Examples for GPT-$3$?

GPT-$3$ has attracted lots of attention due to its superior performance across a wide range of NLP tasks, especially with its powerful and versatile in-context few-shot learning ability. Despite its success, we found that the empirical results of GPT-$3$ depend heavily on the choice of in-context examples. In this work, we investigate whether there are more effective strategies for judiciously selecting in-context examples (relative to random sampling) that better leverage GPT-$3$'s few-shot capabilities. Inspired by the recent success of leveraging a retrieval module to augment large-scale neural network models, we propose to retrieve examples that are semantically-similar to a test sample to formulate its corresponding prompt. Intuitively, the in-context examples selected with such a strategy may serve as more informative inputs to unleash GPT-$3$'s extensive knowledge. We evaluate the proposed approach on several natural language understanding and generation benchmarks, where the retrieval-based prompt selection approach consistently outperforms the random baseline. Moreover, it is observed that the sentence encoders fine-tuned on task-related datasets yield even more helpful retrieval results. Notably, significant gains are observed on tasks such as table-to-text generation (41.9% on the ToTTo dataset) and open-domain question answering (45.5% on the NQ dataset). We hope our investigation could help understand the behaviors of GPT-$3$ and large-scale pre-trained LMs in general and enhance their few-shot capabilities.

SDA: Improving Text Generation with Self Data Augmentation

Data augmentation has been widely used to improve deep neural networks in many research fields, such as computer vision. However, less work has been done in the context of text, partially due to its discrete nature and the complexity of natural languages. In this paper, we propose to improve the standard maximum likelihood estimation (MLE) paradigm by incorporating a self-imitation-learning phase for automatic data augmentation. Unlike most existing sentence-level augmentation strategies, which are only applied to specific models, our method is more general and could be easily adapted to any MLE-based training procedure. In addition, our framework allows task-specific evaluation metrics to be designed to flexibly control the generated sentences, for example, in terms of controlling vocabulary usage and avoiding nontrivial repetitions. Extensive experimental results demonstrate the superiority of our method on two synthetic and several standard real datasets, significantly improving related baselines.

Narrative Incoherence Detection

Motivated by the increasing popularity of intelligent editing assistant, we introduce and investigate the task of narrative incoherence detection: Given a (corrupted) long-form narrative, decide whether there exists some semantic discrepancy in the narrative flow. Specifically, we focus on the missing sentence and incoherent sentence detection. Despite its simple setup, this task is challenging as the model needs to understand and analyze a multi-sentence narrative text, and make decisions at the sentence level. As an initial step towards this task, we implement several baselines either directly analyzing the raw text (\textit{token-level}) or analyzing learned sentence representations (\textit{sentence-level}). We observe that while token-level modeling enjoys greater expressive power and hence better performance, sentence-level modeling possesses an advantage in efficiency and flexibility. With pre-training on large-scale data and cycle-consistent sentence embedding, our extended sentence-level model can achieve comparable detection accuracy to the token-level model. As a by-product, such a strategy enables simultaneous incoherence detection and infilling/modification suggestions.