To reduce a model size but retain performance, we often rely on knowledge distillation (KD) which transfers knowledge from a large teacher'' model to a smaller student'' model. However, KD on multimodal datasets such as vision-language tasks is relat
ively unexplored, and digesting multimodal information is challenging since different modalities present different types of information. In this paper, we perform a large-scale empirical study to investigate the importance and effects of each modality in knowledge distillation. Furthermore, we introduce a multimodal knowledge distillation framework, modality-specific distillation (MSD), to transfer knowledge from a teacher on multimodal tasks by learning the teacher's behavior within each modality. The idea aims at mimicking a teacher's modality-specific predictions by introducing auxiliary loss terms for each modality. Furthermore, because each modality has different saliency for predictions, we define saliency scores for each modality and investigate saliency-based weighting schemes for the auxiliary losses. We further study a weight learning approach to learn the optimal weights on these loss terms. In our empirical analysis, we examine the saliency of each modality in KD, demonstrate the effectiveness of the weighting scheme in MSD, and show that it achieves better performance than KD on four multimodal datasets.
We study multilingual AMR parsing from the perspective of knowledge distillation, where the aim is to learn and improve a multilingual AMR parser by using an existing English parser as its teacher. We constrain our exploration in a strict multilingua
l setting: there is but one model to parse all different languages including English. We identify that noisy input and precise output are the key to successful distillation. Together with extensive pre-training, we obtain an AMR parser whose performances surpass all previously published results on four different foreign languages, including German, Spanish, Italian, and Chinese, by large margins (up to 18.8 Smatch points on Chinese and on average 11.3 Smatch points). Our parser also achieves comparable performance on English to the latest state-of-the-art English-only parser.
Intermediate layer matching is shown as an effective approach for improving knowledge distillation (KD). However, this technique applies matching in the hidden spaces of two different networks (i.e. student and teacher), which lacks clear interpretab
ility. Moreover, intermediate layer KD cannot easily deal with other problems such as layer mapping search and architecture mismatch (i.e. it requires the teacher and student to be of the same model type). To tackle the aforementioned problems all together, we propose Universal-KD to match intermediate layers of the teacher and the student in the output space (by adding pseudo classifiers on intermediate layers) via the attention-based layer projection. By doing this, our unified approach has three merits: (i) it can be flexibly combined with current intermediate layer distillation techniques to improve their results (ii) the pseudo classifiers of the teacher can be deployed instead of extra expensive teacher assistant networks to address the capacity gap problem in KD which is a common issue when the gap between the size of the teacher and student networks becomes too large; (iii) it can be used in cross-architecture intermediate layer KD. We did comprehensive experiments in distilling BERT-base into BERT-4, RoBERTa-large into DistilRoBERTa and BERT-base into CNN and LSTM-based models. Results on the GLUE tasks show that our approach is able to outperform other KD techniques.
Pretrained transformer-based encoders such as BERT have been demonstrated to achieve state-of-the-art performance on numerous NLP tasks. Despite their success, BERT style encoders are large in size and have high latency during inference (especially o
n CPU machines) which make them unappealing for many online applications. Recently introduced compression and distillation methods have provided effective ways to alleviate this shortcoming. However, the focus of these works has been mainly on monolingual encoders. Motivated by recent successes in zero-shot cross-lingual transfer learning using multilingual pretrained encoders such as mBERT, we evaluate the effectiveness of Knowledge Distillation (KD) both during pretraining stage and during fine-tuning stage on multilingual BERT models. We demonstrate that in contradiction to the previous observation in the case of monolingual distillation, in multilingual settings, distillation during pretraining is more effective than distillation during fine-tuning for zero-shot transfer learning. Moreover, we observe that distillation during fine-tuning may hurt zero-shot cross-lingual performance. Finally, we demonstrate that distilling a larger model (BERT Large) results in the strongest distilled model that performs best both on the source language as well as target languages in zero-shot settings.
Knowledge Distillation (KD) offers a natural way to reduce the latency and memory/energy usage of massive pretrained models that have come to dominate Natural Language Processing (NLP) in recent years. While numerous sophisticated variants of KD algo
rithms have been proposed for NLP applications, the key factors underpinning the optimal distillation performance are often confounded and remain unclear. We aim to identify how different components in the KD pipeline affect the resulting performance and how much the optimal KD pipeline varies across different datasets/tasks, such as the data augmentation policy, the loss function, and the intermediate representation for transferring the knowledge between teacher and student. To tease apart their effects, we propose Distiller, a meta KD framework that systematically combines a broad range of techniques across different stages of the KD pipeline, which enables us to quantify each component's contribution. Within Distiller, we unify commonly used objectives for distillation of intermediate representations under a universal mutual information (MI) objective and propose a class of MI-objective functions with better bias/variance trade-off for estimating the MI between the teacher and the student. On a diverse set of NLP datasets, the best Distiller configurations are identified via large-scale hyper-parameter optimization. Our experiments reveal the following: 1) the approach used to distill the intermediate representations is the most important factor in KD performance, 2) among different objectives for intermediate distillation, MI-performs the best, and 3) data augmentation provides a large boost for small training datasets or small student networks. Moreover, we find that different datasets/tasks prefer different KD algorithms, and thus propose a simple AutoDistiller algorithm that can recommend a good KD pipeline for a new dataset.
While neural networks produce state-of-the- art performance in several NLP tasks, they generally depend heavily on lexicalized information, which transfer poorly between domains. Previous works have proposed delexicalization as a form of knowledge di
stillation to reduce the dependency on such lexical artifacts. However, a critical unsolved issue that remains is how much delexicalization to apply: a little helps reduce overfitting, but too much discards useful information. We propose Group Learning, a knowledge and model distillation approach for fact verification in which multiple student models have access to different delexicalized views of the data, but are encouraged to learn from each other through pair-wise consistency losses. In several cross-domain experiments between the FEVER and FNC fact verification datasets, we show that our approach learns the best delexicalization strategy for the given training dataset, and outperforms state-of-the-art classifiers that rely on the original data.
We present an efficient training approach to text retrieval with dense representations that applies knowledge distillation using the ColBERT late-interaction ranking model. Specifically, we propose to transfer the knowledge from a bi-encoder teacher
to a student by distilling knowledge from ColBERT's expressive MaxSim operator into a simple dot product. The advantage of the bi-encoder teacher--student setup is that we can efficiently add in-batch negatives during knowledge distillation, enabling richer interactions between teacher and student models. In addition, using ColBERT as the teacher reduces training cost compared to a full cross-encoder. Experiments on the MS MARCO passage and document ranking tasks and data from the TREC 2019 Deep Learning Track demonstrate that our approach helps models learn robust representations for dense retrieval effectively and efficiently.
In this paper we apply self-knowledge distillation to text summarization which we argue can alleviate problems with maximum-likelihood training on single reference and noisy datasets. Instead of relying on one-hot annotation labels, our student summa
rization model is trained with guidance from a teacher which generates smoothed labels to help regularize training. Furthermore, to better model uncertainty during training, we introduce multiple noise signals for both teacher and student models. We demonstrate experimentally on three benchmarks that our framework boosts the performance of both pretrained and non-pretrained summarizers achieving state-of-the-art results.
Large neural networks are impractical to deploy on mobile devices due to their heavy computational cost and slow inference. Knowledge distillation (KD) is a technique to reduce the model size while retaining performance by transferring knowledge from
a large teacher'' model to a smaller student'' model. However, KD on multimodal datasets such as vision-language datasets is relatively unexplored and digesting such multimodal information is challenging since different modalities present different types of information. In this paper, we propose modality-specific distillation (MSD) to effectively transfer knowledge from a teacher on multimodal datasets. Existing KD approaches can be applied to multimodal setup, but a student doesn't have access to modality-specific predictions. Our idea aims at mimicking a teacher's modality-specific predictions by introducing an auxiliary loss term for each modality. Because each modality has different importance for predictions, we also propose weighting approaches for the auxiliary losses; a meta-learning approach to learn the optimal weights on these loss terms. In our experiments, we demonstrate the effectiveness of our MSD and the weighting scheme and show that it achieves better performance than KD.
In most of neural machine translation distillation or stealing scenarios, the highest-scoring hypothesis of the target model (teacher) is used to train a new model (student). If reference translations are also available, then better hypotheses (with
respect to the references) can be oversampled and poor hypotheses either removed or undersampled. This paper explores the sampling method landscape (pruning, hypothesis oversampling and undersampling, deduplication and their combination) with English to Czech and English to German MT models using standard MT evaluation metrics. We show that careful oversampling and combination with the original data leads to better performance when compared to training only on the original or synthesized data or their direct combination.
