Recent work has questioned the importance of the Transformer's multi-headed attention for achieving high translation quality. We push further in this direction by developing a "hardcoded" attention variant without any learned parameters. Surprisingly, replacing all learned self-attention heads in the encoder and decoder with fixed, input-agnostic Gaussian distributions minimally impacts BLEU scores across four different language pairs. However, additionally hard-coding cross attention (which connects the decoder to the encoder) significantly lowers BLEU, suggesting that it is more important than self-attention. Much of this BLEU drop can be recovered by adding just a single learned cross attention head to an otherwise hard-coded Transformer. Taken as a whole, our results offer insight into which components of the Transformer are actually important, which we hope will guide future work into the development of simpler and more efficient attention-based models.

Until recently, summarization evaluations compared systems that produce summaries of the same target length. Neural approaches to summarization however have done away with length requirements. Here we present detailed experiments demonstrating that summaries of different length produced by the same system have a clear non-linear pattern of quality as measured by ROUGE F1 scores: initially steeply improving with summary length, then starting to gradually decline. Neural models produce summaries of different length, possibly confounding improvements of summarization techniques with potentially spurious learning of optimal summary length. We propose a new evaluation method where ROUGE scores are normalized by those of a random system producing summaries of the same length. We reanalyze a number of recently reported results and show that some negative results are in fact reports of system improvement once differences in length are taken into account. Finally, we present a small-scale human evaluation showing a similar trend of perceived quality increase with summary length, calling for the need of similar normalization in reporting human scores.

ROUGE is widely used to automatically evaluate summarization systems. However, ROUGE measures semantic overlap between a system summary and a human reference on word-string level, much at odds with the contemporary treatment of semantic meaning. Here we present a suite of experiments on using distributed representations for evaluating summarizers, both in reference-based and in reference-free setting. Our experimental results show that the max value over each dimension of the summary ELMo word embeddings is a good representation that results in high correlation with human ratings. Averaging the cosine similarity of all encoders we tested yields high correlation with manual scores in reference-free setting. The distributed representations outperform ROUGE in recent corpora for abstractive news summarization but are less good on older test data and systems.

Language models are generally trained on short, truncated input sequences, which limits their ability to use discourse-level information present in long-range context to improve their predictions. Recent efforts to improve the efficiency of self-attention have led to a proliferation of long-range Transformer language models, which can process much longer sequences than models of the past. However, the ways in which such models take advantage of the longrange context remain unclear. In this paper, we perform a fine-grained analysis of two longrange Transformer language models (including the Routing Transformer, which achieves state-of-the-art perplexity on the PG-19 longsequence LM benchmark dataset) that accept input sequences of up to 8K tokens. Our results reveal that providing long-range context (i.e., beyond the previous 2K tokens) to these models only improves their predictions on a small set of tokens (e.g., those that can be copied from the distant context) and does not help at all for sentence-level prediction tasks. Finally, we discover that PG-19 contains a variety of different document types and domains, and that long-range context helps most for literary novels (as opposed to textbooks or magazines).

No abstract

No abstract

The discrepancy between maximum likelihood estimation (MLE) and task measures such as BLEU score has been studied before for autoregressive neural machine translation (NMT) and resulted in alternative training algorithms (Ranzato et al., 2016;Norouzi et al., 2016;Shen et al., 2016;Wu et al., 2018). However, MLE training remains the de facto approach for autoregressive NMT because of its computational efficiency and stability. Despite this mismatch between the training objective and task measure, we notice that the samples drawn from an MLE-based trained NMT support the desired distribution -there are samples with much higher BLEU score comparing to the beam decoding output. To benefit from this observation, we train an energybased model to mimic the behavior of the task measure (i.e., the energy-based model assigns lower energy to samples with higher BLEU score), which is resulted in a re-ranking algorithm based on the samples drawn from NMT: energy-based re-ranking (EBR). We use both marginal energy models (over target sentence) and joint energy models (over both source and target sentences). Our EBR with the joint energy model consistently improves the performance of the Transformer-based NMT: +3.7 BLEU points on IWSLT'14 German-English, +3.37 BELU points on Sinhala-English, +1.4 BLEU points on WMT'16 English-German tasks.

Traditional single image super-resolution (SISR) methods that focus on solving single and uniform degradation (i.e., bicubic down-sampling), typically suffer from poor performance when applied into real-world low-resolution (LR) images due to the complicated realistic degradations. The key to solving this more challenging real image super-resolution (RealSR) problem lies in learning feature representations that are both informative and content-aware. In this paper, we propose a Omni-frequency Region-adaptive Network (OR-Net) to address both challenges, here we call features of all low, middle and high frequencies omni-frequency features. Specifically, we start from the frequency perspective and design a Frequency Decomposition (FD) module to separate different frequency components to comprehensively compensate the information lost for real LR image. Then, considering the different regions of real LR image have different frequency information lost, we further design a Region-adaptive Frequency Aggregation (RFA) module by leveraging dynamic convolution and spatial attention to adaptively restore frequency components for different regions. The extensive experiments endorse the high-efficient, effective, and scenario-agnostic nature of our OR-Net for RealSR.