We present Corgi, a novel method for text-to-image generation. Corgi is based on our proposed shifted diffusion model, which achieves better image embedding generation from input text. Unlike the baseline diffusion model used in DALL-E 2, our method seamlessly encodes prior knowledge of the pre-trained CLIP model in its diffusion process by designing a new initialization distribution and a new transition step of the diffusion. Compared to the strong DALL-E 2 baseline, our method performs better in generating image embedding from the text in terms of both efficiency and effectiveness, resulting in better text-to-image generation. Extensive large-scale experiments are conducted and evaluated in terms of both quantitative measures and human evaluation, indicating a stronger generation ability of our method compared to existing ones. Furthermore, our model enables semi-supervised and language-free training for text-to-image generation, where only part or none of the images in the training dataset have an associated caption. Trained with only 1.7% of the images being captioned, our semi-supervised model obtains FID results comparable to DALL-E 2 on zero-shot text-to-image generation evaluated on MS-COCO. Corgi also achieves new state-of-the-art results across different datasets on downstream language-free text-to-image generation tasks, outperforming the previous method, Lafite, by a large margin.
Diffusion models were initially developed for text-to-image generation and are now being utilized to generate high quality synthetic images. Preceded by GANs, diffusion models have shown impressive results using various evaluation metrics. However, commonly used metrics such as FID and IS are not suitable for determining whether diffusion models are simply reproducing the training images. Here we train StyleGAN and diffusion models, using BRATS20 and BRATS21 datasets, to synthesize brain tumor images, and measure the correlation between the synthetic images and all training images. Our results show that diffusion models are much more likely to memorize the training images, especially for small datasets. Researchers should be careful when using diffusion models for medical imaging, if the final goal is to share the synthetic images.
The creation of relevance assessments by human assessors (often nowadays crowdworkers) is a vital step when building IR test collections. Prior works have investigated assessor quality & behaviour, though into the impact of a document's presentation modality on assessor efficiency and effectiveness. Given the rise of voice-based interfaces, we investigate whether it is feasible for assessors to judge the relevance of text documents via a voice-based interface. We ran a user study (n = 49) on a crowdsourcing platform where participants judged the relevance of short and long documents sampled from the TREC Deep Learning corpus-presented to them either in the text or voice modality. We found that: (i) participants are equally accurate in their judgements across both the text and voice modality; (ii) with increased document length it takes participants significantly longer (for documents of length > 120 words it takes almost twice as much time) to make relevance judgements in the voice condition; and (iii) the ability of assessors to ignore stimuli that are not relevant (i.e., inhibition) impacts the assessment quality in the voice modality-assessors with higher inhibition are significantly more accurate than those with lower inhibition. Our results indicate that we can reliably leverage the voice modality as a means to effectively collect relevance labels from crowdworkers.
Is it possible to leverage large scale raw and raw parallel corpora to build a general learned metric? Existing learned metrics have gaps to human judgements, are model-dependent or are limited to the domains or tasks where human ratings are available. In this paper, we propose SEScore2, a model-based metric pretrained over million-scale synthetic dataset constructed by our novel retrieval augmented data synthesis pipeline. SEScore2 achieves high correlation to human judgements without any human rating supervisions. Importantly, our unsupervised SEScore2 can outperform supervised metrics, which are trained on the News human ratings, at the TED domain. We evaluate SEScore2 over four text generation tasks across three languages. SEScore2 outperforms all prior unsupervised evaluation metrics in machine translation, speech translation, data-to-text and dialogue generation, with average Kendall improvements 0.158. SEScore2 even outperforms SOTA supervised BLEURT at data-to-text, dialogue generation and overall correlation.
Self-supervised speech pre-training empowers the model with the contextual structure inherent in the speech signal while self-supervised text pre-training empowers the model with linguistic information. Both of them are beneficial for downstream speech tasks such as ASR. However, the distinct pre-training objectives make it challenging to jointly optimize the speech and text representation in the same model. To solve this problem, we propose Text-Enhanced Self-Supervised Speech Pre-training (TESSP), aiming to incorporate the linguistic information into speech pre-training. Our model consists of three parts, i.e., a speech encoder, a text encoder and a shared encoder. The model takes unsupervised speech and text data as the input and leverages the common HuBERT and MLM losses respectively. We also propose phoneme up-sampling and representation swapping to enable joint modeling of the speech and text information. Specifically, to fix the length mismatching problem between speech and text data, we phonemize the text sequence and up-sample the phonemes with the alignment information extracted from a small set of supervised data. Moreover, to close the gap between the learned speech and text representations, we swap the text representation with the speech representation extracted by the respective private encoders according to the alignment information. Experiments on the Librispeech dataset shows the proposed TESSP model achieves more than 10% improvement compared with WavLM on the test-clean and test-other sets. We also evaluate our model on the SUPERB benchmark, showing our model has better performance on Phoneme Recognition, Acoustic Speech Recognition and Speech Translation compared with WavLM.
Large language models (LLMs) have demonstrated impressive zero-shot abilities on a variety of open-ended tasks, while recent research has also explored the use of LLMs for multi-modal generation. In this study, we introduce mPLUG-Owl, a novel training paradigm that equips LLMs with multi-modal abilities through modularized learning of foundation LLM, a visual knowledge module, and a visual abstractor module. This approach can support multiple modalities and facilitate diverse unimodal and multimodal abilities through modality collaboration. The training paradigm of mPLUG-Owl involves a two-stage method for aligning image and text, which learns visual knowledge with the assistance of LLM while maintaining and even improving the generation abilities of LLM. In the first stage, the visual knowledge module and abstractor module are trained with a frozen LLM module to align the image and text. In the second stage, language-only and multi-modal supervised datasets are used to jointly fine-tune a low-rank adaption (LoRA) module on LLM and the abstractor module by freezing the visual knowledge module. We carefully build a visually-related instruction evaluation set OwlEval. Experimental results show that our model outperforms existing multi-modal models, demonstrating mPLUG-Owl's impressive instruction and visual understanding ability, multi-turn conversation ability, and knowledge reasoning ability. Besides, we observe some unexpected and exciting abilities such as multi-image correlation and scene text understanding, which makes it possible to leverage it for harder real scenarios, such as vision-only document comprehension. Our code, pre-trained model, instruction-tuned models, and evaluation set are available at https://github.com/X-PLUG/mPLUG-Owl. The online demo is available at https://www.modelscope.cn/studios/damo/mPLUG-Owl.
Text Summarization is recognised as one of the NLP downstream tasks and it has been extensively investigated in recent years. It can assist people with perceiving the information rapidly from the Internet, including news articles, social posts, videos, etc. Most existing research works attempt to develop summarization models to produce a better output. However, advent limitations of most existing models emerge, including unfaithfulness and factual errors. In this paper, we propose a novel model, named as Knowledge-aware Abstractive Text Summarization, which leverages the advantages offered by Knowledge Graph to enhance the standard Seq2Seq model. On top of that, the Knowledge Graph triplets are extracted from the source text and utilised to provide keywords with relational information, producing coherent and factually errorless summaries. We conduct extensive experiments by using real-world data sets. The results reveal that the proposed framework can effectively utilise the information from Knowledge Graph and significantly reduce the factual errors in the summary.
Online resources such as WikiHow compile a wide range of scripts for performing everyday tasks, which can assist models in learning to reason about procedures. However, the scripts are always presented in a linear manner, which does not reflect the flexibility displayed by people executing tasks in real life. For example, in the CrossTask Dataset, 64.5% of consecutive step pairs are also observed in the reverse order, suggesting their ordering is not fixed. In addition, each step has an average of 2.56 frequent next steps, demonstrating "branching". In this paper, we propose the new challenging task of non-sequential graph script induction, aiming to capture optional and interchangeable steps in procedural planning. To automate the induction of such graph scripts for given tasks, we propose to take advantage of loosely aligned videos of people performing the tasks. In particular, we design a multimodal framework to ground procedural videos to WikiHow textual steps and thus transform each video into an observed step path on the latent ground truth graph script. This key transformation enables us to train a script knowledge model capable of both generating explicit graph scripts for learnt tasks and predicting future steps given a partial step sequence. Our best model outperforms the strongest pure text/vision baselines by 17.52% absolute gains on F1@3 for next step prediction and 13.8% absolute gains on Acc@1 for partial sequence completion. Human evaluation shows our model outperforming the WikiHow linear baseline by 48.76% absolute gains in capturing sequential and non-sequential step relationships.
The influx of deep learning (DL) techniques into the field of survival analysis in recent years, coupled with the increasing availability of high-dimensional omics data and unstructured data like images or text, has led to substantial methodological progress; for instance, learning from such high-dimensional or unstructured data. Numerous modern DL-based survival methods have been developed since the mid-2010s; however, they often address only a small subset of scenarios in the time-to-event data setting - e.g., single-risk right-censored survival tasks - and neglect to incorporate more complex (and common) settings. Partially, this is due to a lack of exchange between experts in the respective fields. In this work, we provide a comprehensive systematic review of DL-based methods for time-to-event analysis, characterizing them according to both survival- and DL-related attributes. In doing so, we hope to provide a helpful overview to practitioners who are interested in DL techniques applicable to their specific use case as well as to enable researchers from both fields to identify directions for future investigation. We provide a detailed characterization of the methods included in this review as an open-source, interactive table: https://survival-org.github.io/DL4Survival. As this research area is advancing rapidly, we encourage the research community to contribute to keeping the information up to date.
LLMs have shown impressive few-shot performance across many tasks. However, they still struggle when it comes to reliably generating complex output structures, such as those required for information extraction. This limitation stems from the fact that LLMs, without fine-tuning, tend to generate free text rather than structures precisely following a specific grammar. In this work, we propose to enrich the decoding with formal grammar constraints. More concretely, given Context-Free Grammar(CFG), our framework ensures that the token generated in each decoding step would lead to a valid continuation compliant with the grammar production rules. This process guarantees the generation of valid sequences. Importantly, our framework can be readily combined with any CFG or decoding algorithm. We demonstrate that the outputs of many NLP tasks can be represented as formal languages, making them suitable for direct use in our framework. We conducted experiments with two challenging tasks involving large alphabets in their grammar (Wikidata entities and relations): information extraction and entity disambiguation. Our results with LLaMA models indicate that grammar-constrained decoding substantially outperforms unconstrained decoding and even competes with task-specific fine-tuned models. These findings suggest that integrating grammar-based constraints during decoding holds great promise in making LLMs reliably produce structured outputs, especially in setting where training data is scarce and fine-tuning is expensive.