The increasing reliance on Large Language Models (LLMs) across academia and industry necessitates a comprehensive understanding of their robustness to prompts. In response to this vital need, we introduce PromptRobust, a robustness benchmark designed to measure LLMs' resilience to adversarial prompts. This study uses a plethora of adversarial textual attacks targeting prompts across multiple levels: character, word, sentence, and semantic. The adversarial prompts, crafted to mimic plausible user errors like typos or synonyms, aim to evaluate how slight deviations can affect LLM outcomes while maintaining semantic integrity. These prompts are then employed in diverse tasks including sentiment analysis, natural language inference, reading comprehension, machine translation, and math problem-solving. Our study generates 4,788 adversarial prompts, meticulously evaluated over 8 tasks and 13 datasets. Our findings demonstrate that contemporary LLMs are not robust to adversarial prompts. Furthermore, we present a comprehensive analysis to understand the mystery behind prompt robustness and its transferability. We then offer insightful robustness analysis and pragmatic recommendations for prompt composition, beneficial to both researchers and everyday users.
Word error rate (WER) and character error rate (CER) are standard metrics in Speech Recognition (ASR), but one problem has always been alternative spellings: If one's system transcribes adviser whereas the ground truth has advisor, this will count as an error even though the two spellings really represent the same word. Japanese is notorious for ``lacking orthography'': most words can be spelled in multiple ways, presenting a problem for accurate ASR evaluation. In this paper we propose a new lenient evaluation metric as a more defensible CER measure for Japanese ASR. We create a lattice of plausible respellings of the reference transcription, using a combination of lexical resources, a Japanese text-processing system, and a neural machine translation model for reconstructing kanji from hiragana or katakana. In a manual evaluation, raters rated 95.4% of the proposed spelling variants as plausible. ASR results show that our method, which does not penalize the system for choosing a valid alternate spelling of a word, affords a 2.4%-3.1% absolute reduction in CER depending on the task.
Prompt engineering is an essential technique for enhancing the abilities of large language models (LLMs) by providing explicit and specific instructions. It enables LLMs to excel in various tasks, such as arithmetic reasoning, question answering, summarization, relation extraction, machine translation, and sentiment analysis. Researchers have been actively exploring different prompt engineering strategies, such as Chain of Thought (CoT), Zero-CoT, and In-context learning. However, an unresolved problem arises from the fact that current approaches lack a solid mathematical solution for determining optimal prompts. To address this issue in prompt engineering, we propose a new and effective approach called Prompt Space. Our methodology utilizes text embeddings to obtain basis vectors by matrix decomposition, and then constructs a space for representing all prompts. Prompt Space significantly outperforms state-of-the-art prompt paradigms on ten public reasoning benchmarks. Notably, without the help of the CoT method and the prompt "Let's think step by step", Prompt Space shows superior performance over the few-shot method. Overall, our approach provides a robust and effective mathematical framework for selecting simple and effective prompts. This advancement marks a significant step towards improving prompt engineering for a wide variety of applications in LLMs. Our code is publicly available at \textcolor{blue}{\url{https://github.com/YouBLEI/Prompt-Space}}
Quality Estimation (QE) is the task of evaluating the quality of a translation when reference translation is not available. The goal of QE aligns with the task of corpus filtering, where we assign the quality score to the sentence pairs present in the pseudo-parallel corpus. We propose a Quality Estimation based Filtering approach to extract high-quality parallel data from the pseudo-parallel corpus. To the best of our knowledge, this is a novel adaptation of the QE framework to extract quality parallel corpus from the pseudo-parallel corpus. By training with this filtered corpus, we observe an improvement in the Machine Translation (MT) system's performance by up to 1.8 BLEU points, for English-Marathi, Chinese-English, and Hindi-Bengali language pairs, over the baseline model. The baseline model is the one that is trained on the whole pseudo-parallel corpus. Our Few-shot QE model transfer learned from the English-Marathi QE model and fine-tuned on only 500 Hindi-Bengali training instances, shows an improvement of up to 0.6 BLEU points for Hindi-Bengali language pair, compared to the baseline model. This demonstrates the promise of transfer learning in the setting under discussion. QE systems typically require in the order of (7K-25K) of training data. Our Hindi-Bengali QE is trained on only 500 instances of training that is 1/40th of the normal requirement and achieves comparable performance. All the scripts and datasets utilized in this study will be publicly available.
A major challenge in the field of Text Generation is evaluation: Human evaluations are cost-intensive, and automated metrics often display considerable disagreement with human judgments. In this paper, we propose a statistical model of Text Generation evaluation that accounts for the error-proneness of automated metrics when used to generate preference rankings between system outputs. We show that existing automated metrics are generally over-confident in assigning significant differences between systems in this setting. However, our model enables an efficient combination of human and automated ratings to remedy the error-proneness of the automated metrics. We show that using this combination, we only require about 50% of the human annotations typically used in evaluations to arrive at robust and statistically significant results while yielding the same evaluation outcome as the pure human evaluation in 95% of cases. We showcase the benefits of approach for three text generation tasks: dialogue systems, machine translation, and text summarization.
Ternary and binary neural networks enable multiplication-free computation and promise multiple orders of magnitude efficiency gains over full-precision networks if implemented on specialized hardware. However, since both the parameter and the output space are highly discretized, such networks have proven very difficult to optimize. The difficulties are compounded for the class of transformer text generation models due to the sensitivity of the attention operation to quantization and the noise-compounding effects of autoregressive decoding in the high-cardinality output space. We approach the problem with a mix of statistics-based quantization for the weights and elastic quantization of the activations and demonstrate the first ternary and binary transformer models on the downstream tasks of summarization and machine translation. Our ternary BART base achieves an R1 score of 41 on the CNN/DailyMail benchmark, which is merely 3.9 points behind the full model while being 16x more efficient. Our binary model, while less accurate, achieves a highly non-trivial score of 35.6. For machine translation, we achieved BLEU scores of 21.7 and 17.6 on the WMT16 En-Ro benchmark, compared with a full precision mBART model score of 26.8. We also compare our approach in the 8-bit activation setting, where our ternary and even binary weight models can match or outperform the best existing 8-bit weight models in the literature. Our code and models are available at: https://github.com/facebookresearch/Ternary_Binary_Transformer
Text simplification aims to make the text easier to understand by applying rewriting transformations. There has been very little research on Chinese text simplification for a long time. The lack of generic evaluation data is an essential reason for this phenomenon. In this paper, we introduce MCTS, a multi-reference Chinese text simplification dataset. We describe the annotation process of the dataset and provide a detailed analysis. Furthermore, we evaluate the performance of several unsupervised methods and advanced large language models. We additionally provide Chinese text simplification parallel data that can be used for training, acquired by utilizing machine translation and English text simplification. We hope to build a basic understanding of Chinese text simplification through the foundational work and provide references for future research. All of the code and data are released at https://github.com/blcuicall/mcts/.
While Neural Machine Translation(NMT) has achieved great progress in recent years, it still suffers from inaccurate translation of entities (e.g., person/organization name, location), due to the lack of entity training instances. When we humans encounter an unknown entity during translation, we usually first look up in a dictionary and then organize the entity translation together with the translations of other parts to form a smooth target sentence. Inspired by this translation process, we propose an Extract-and-Attend approach to enhance entity translation in NMT, where the translation candidates of source entities are first extracted from a dictionary and then attended to by the NMT model to generate the target sentence. Specifically, the translation candidates are extracted by first detecting the entities in a source sentence and then translating the entities through looking up in a dictionary. Then, the extracted candidates are added as a prefix of the decoder input to be attended to by the decoder when generating the target sentence through self-attention. Experiments conducted on En-Zh and En-Ru demonstrate that the proposed method is effective on improving both the translation accuracy of entities and the overall translation quality, with up to 35% reduction on entity error rate and 0.85 gain on BLEU and 13.8 gain on COMET.
This paper presents a new approach for assessing uncertainty in machine translation by simultaneously evaluating translation quality and providing a reliable confidence score. Our approach utilizes conformal predictive distributions to produce prediction intervals with guaranteed coverage, meaning that for any given significance level $\epsilon$, we can expect the true quality score of a translation to fall out of the interval at a rate of $1-\epsilon$. In this paper, we demonstrate how our method outperforms a simple, but effective baseline on six different language pairs in terms of coverage and sharpness. Furthermore, we validate that our approach requires the data exchangeability assumption to hold for optimal performance.
Subword tokenization is the de facto standard for tokenization in neural language models and machine translation systems. Three advantages are frequently cited in favor of subwords: shorter encoding of frequent tokens, compositionality of subwords, and ability to deal with unknown words. As their relative importance is not entirely clear yet, we propose a tokenization approach that enables us to separate frequency (the first advantage) from compositionality. The approach uses Huffman coding to tokenize words, by order of frequency, using a fixed amount of symbols. Experiments with CS-DE, EN-FR and EN-DE NMT show that frequency alone accounts for 90%-95% of the scores reached by BPE, hence compositionality has less importance than previously thought.
This paper describes the submission of the UPC Machine Translation group to the IWSLT 2023 Offline Speech Translation task. Our Speech Translation systems utilize foundation models for speech (wav2vec 2.0) and text (mBART50). We incorporate a Siamese pretraining step of the speech and text encoders with CTC and Optimal Transport, to adapt the speech representations to the space of the text model, thus maximizing transfer learning from MT. After this pretraining, we fine-tune our system end-to-end on ST, with Cross Entropy and Knowledge Distillation. Apart from the available ST corpora, we create synthetic data with SegAugment to better adapt our models to the custom segmentations of the IWSLT test sets. Our best single model obtains 31.2 BLEU points on MuST-C tst-COMMON, 29.8 points on IWLST.tst2020 and 33.4 points on the newly released IWSLT.ACLdev2023.
Back Translation (BT) is widely used in the field of machine translation, as it has been proved effective for enhancing translation quality. However, BT mainly improves the translation of inputs that share a similar style (to be more specific, translation-like inputs), since the source side of BT data is machine-translated. For natural inputs, BT brings only slight improvements and sometimes even adverse effects. To address this issue, we propose Text Style Transfer Back Translation (TST BT), which uses a style transfer model to modify the source side of BT data. By making the style of source-side text more natural, we aim to improve the translation of natural inputs. Our experiments on various language pairs, including both high-resource and low-resource ones, demonstrate that TST BT significantly improves translation performance against popular BT benchmarks. In addition, TST BT is proved to be effective in domain adaptation so this strategy can be regarded as a general data augmentation method. Our training code and text style transfer model are open-sourced.
This paper investigates the impact of data volume and the use of similar languages on transfer learning in a machine translation task. We find out that having more data generally leads to better performance, as it allows the model to learn more patterns and generalizations from the data. However, related languages can also be particularly effective when there is limited data available for a specific language pair, as the model can leverage the similarities between the languages to improve performance. To demonstrate, we fine-tune mBART model for a Polish-English translation task using the OPUS-100 dataset. We evaluate the performance of the model under various transfer learning configurations, including different transfer source languages and different shot levels for Polish, and report the results. Our experiments show that a combination of related languages and larger amounts of data outperforms the model trained on related languages or larger amounts of data alone. Additionally, we show the importance of related languages in zero-shot and few-shot configurations.
We tackle the task of automatically discriminating between human and machine translations. As opposed to most previous work, we perform experiments in a multilingual setting, considering multiple languages and multilingual pretrained language models. We show that a classifier trained on parallel data with a single source language (in our case German-English) can still perform well on English translations that come from different source languages, even when the machine translations were produced by other systems than the one it was trained on. Additionally, we demonstrate that incorporating the source text in the input of a multilingual classifier improves (i) its accuracy and (ii) its robustness on cross-system evaluation, compared to a monolingual classifier. Furthermore, we find that using training data from multiple source languages (German, Russian, and Chinese) tends to improve the accuracy of both monolingual and multilingual classifiers. Finally, we show that bilingual classifiers and classifiers trained on multiple source languages benefit from being trained on longer text sequences, rather than on sentences.
In recent years machine translation has become very successful for high-resource language pairs. This has also sparked new interest in research on the automatic translation of low-resource languages, including Indigenous languages. However, the latter are deeply related to the ethnic and cultural groups that speak (or used to speak) them. The data collection, modeling and deploying machine translation systems thus result in new ethical questions that must be addressed. Motivated by this, we first survey the existing literature on ethical considerations for the documentation, translation, and general natural language processing for Indigenous languages. Afterward, we conduct and analyze an interview study to shed light on the positions of community leaders, teachers, and language activists regarding ethical concerns for the automatic translation of their languages. Our results show that the inclusion, at different degrees, of native speakers and community members is vital to performing better and more ethical research on Indigenous languages.
We propose a genetic algorithm (GA) based method for modifying n-best lists produced by a machine translation (MT) system. Our method offers an innovative approach to improving MT quality and identifying weaknesses in evaluation metrics. Using common GA operations (mutation and crossover) on a list of hypotheses in combination with a fitness function (an arbitrary MT metric), we obtain novel and diverse outputs with high metric scores. With a combination of multiple MT metrics as the fitness function, the proposed method leads to an increase in translation quality as measured by other held-out automatic metrics. With a single metric (including popular ones such as COMET) as the fitness function, we find blind spots and flaws in the metric. This allows for an automated search for adversarial examples in an arbitrary metric, without prior assumptions on the form of such example. As a demonstration of the method, we create datasets of adversarial examples and use them to show that reference-free COMET is substantially less robust than the reference-based version.
Pretrained language models (PLMs) have produced substantial improvements in discourse-aware neural machine translation (NMT), for example, improved coherence in spoken language translation. However, the underlying reasons for their strong performance have not been well explained. To bridge this gap, we introduce a probing task to interpret the ability of PLMs to capture discourse relation knowledge. We validate three state-of-the-art PLMs across encoder-, decoder-, and encoder-decoder-based models. The analysis shows that (1) the ability of PLMs on discourse modelling varies from architecture and layer; (2) discourse elements in a text lead to different learning difficulties for PLMs. Besides, we investigate the effects of different PLMs on spoken language translation. Through experiments on IWSLT2017 Chinese-English dataset, we empirically reveal that NMT models initialized from different layers of PLMs exhibit the same trends with the probing task. Our findings are instructive to understand how and when discourse knowledge in PLMs should work for downstream tasks.
We present a large-scale video subtitle translation dataset, BigVideo, to facilitate the study of multi-modality machine translation. Compared with the widely used How2 and VaTeX datasets, BigVideo is more than 10 times larger, consisting of 4.5 million sentence pairs and 9,981 hours of videos. We also introduce two deliberately designed test sets to verify the necessity of visual information: Ambiguous with the presence of ambiguous words, and Unambiguous in which the text context is self-contained for translation. To better model the common semantics shared across texts and videos, we introduce a contrastive learning method in the cross-modal encoder. Extensive experiments on the BigVideo show that: a) Visual information consistently improves the NMT model in terms of BLEU, BLEURT, and COMET on both Ambiguous and Unambiguous test sets. b) Visual information helps disambiguation, compared to the strong text baseline on terminology-targeted scores and human evaluation. Dataset and our implementations are available at https://github.com/DeepLearnXMU/BigVideo-VMT.
Many NLP pipelines split text into sentences as one of the crucial preprocessing steps. Prior sentence segmentation tools either rely on punctuation or require a considerable amount of sentence-segmented training data: both central assumptions might fail when porting sentence segmenters to diverse languages on a massive scale. In this work, we thus introduce a multilingual punctuation-agnostic sentence segmentation method, currently covering 85 languages, trained in a self-supervised fashion on unsegmented text, by making use of newline characters which implicitly perform segmentation into paragraphs. We further propose an approach that adapts our method to the segmentation in a given corpus by using only a small number (64-256) of sentence-segmented examples. The main results indicate that our method outperforms all the prior best sentence-segmentation tools by an average of 6.1% F1 points. Furthermore, we demonstrate that proper sentence segmentation has a point: the use of a (powerful) sentence segmenter makes a considerable difference for a downstream application such as machine translation (MT). By using our method to match sentence segmentation to the segmentation used during training of MT models, we achieve an average improvement of 2.3 BLEU points over the best prior segmentation tool, as well as massive gains over a trivial segmenter that splits text into equally sized blocks.
Although neural-based machine translation evaluation metrics, such as COMET or BLEURT, have achieved strong correlations with human judgements, they are sometimes unreliable in detecting certain phenomena that can be considered as critical errors, such as deviations in entities and numbers. In contrast, traditional evaluation metrics, such as BLEU or chrF, which measure lexical or character overlap between translation hypotheses and human references, have lower correlations with human judgements but are sensitive to such deviations. In this paper, we investigate several ways of combining the two approaches in order to increase robustness of state-of-the-art evaluation methods to translations with critical errors. We show that by using additional information during training, such as sentence-level features and word-level tags, the trained metrics improve their capability to penalize translations with specific troublesome phenomena, which leads to gains in correlation with human judgments and on recent challenge sets on several language pairs.