We noticed that ChatGPT struggled to extract co-reference entities like "her medications" or "her symptoms", which should be annotated in accordance with the 2010 i2b2 annotation guidelines, for coreference identification purposes. After we removed those co-reference entities in the gold standard and re-evaluated the performance of both ChatGPT and GPT-3, we observed modest increases in performance, with ChatGPT achieving an F1 score of 0.628 using Prompt-2 and GPT-3 attaining an F1 score of 0.500 in the relaxed-match criteria. Moreover, we observed a significant degree of randomness in ChatGPT's output. Even when presented with the same prompt and the same input text, it sometimes generated responses with considerable differences in format and content. This phenomenon was particularly prevalent when the input note was lengthy, despite our efforts to minimize input sequence length by limiting it to the HPI section. We anticipate this issue will be addressed when GPT-4 allows much longer text. Although it is not clear whether clinical corpora (and what types of clinical corpora) are used in training ChatGPT, ChatGPT has demonstrated its understanding of the medical text to a certain degree. We believe fine-tuning ChatGPT with domain-specific corpora, assuming OpenAI will provide such an API, will further improve its performance on clinical NLP tasks such as NER in the zero-shot fashion.

Machine learning models have been shown to inherit biases from their training datasets. This can be particularly problematic for vision-language foundation models trained on uncurated datasets scraped from the internet. The biases can be amplified and propagated to downstream applications like zero-shot classifiers and text-to-image generative models. In this study, we propose a general approach for debiasing vision-language foundation models by projecting out biased directions in the text embedding. In particular, we show that debiasing only the text embedding with a calibrated projection matrix suffices to yield robust classifiers and fair generative models. The proposed closed-form solution enables easy integration into large-scale pipelines, and empirical results demonstrate that our approach effectively reduces social bias and spurious correlation in both discriminative and generative vision-language models without the need for additional data or training.

Recently, large language models (LLMs) (e.g., GPT-4) have demonstrated impressive general-purpose task-solving abilities, including the potential to approach recommendation tasks. Along this line of research, this work aims to investigate the capacity of LLMs that act as the ranking model for recommender systems. To conduct our empirical study, we first formalize the recommendation problem as a conditional ranking task, considering sequential interaction histories as conditions and the items retrieved by the candidate generation model as candidates. We adopt a specific prompting approach to solving the ranking task by LLMs: we carefully design the prompting template by including the sequential interaction history, the candidate items, and the ranking instruction. We conduct extensive experiments on two widely-used datasets for recommender systems and derive several key findings for the use of LLMs in recommender systems. We show that LLMs have promising zero-shot ranking abilities, even competitive to or better than conventional recommendation models on candidates retrieved by multiple candidate generators. We also demonstrate that LLMs struggle to perceive the order of historical interactions and can be affected by biases like position bias, while these issues can be alleviated via specially designed prompting and bootstrapping strategies.

In this paper, we present ChatPLUG, a Chinese open-domain dialogue system for digital human applications that instruction finetunes on a wide range of dialogue tasks in a unified internet-augmented format. Different from other open-domain dialogue models that focus on large-scale pre-training and scaling up model size or dialogue corpus, we aim to build a powerful and practical dialogue system for digital human with diverse skills and good multi-task generalization by internet-augmented instruction tuning. To this end, we first conduct large-scale pre-training on both common document corpus and dialogue data with curriculum learning, so as to inject various world knowledge and dialogue abilities into ChatPLUG. Then, we collect a wide range of dialogue tasks spanning diverse features of knowledge, personality, multi-turn memory, and empathy, on which we further instruction tune \modelname via unified natural language instruction templates. External knowledge from an internet search is also used during instruction finetuning for alleviating the problem of knowledge hallucinations. We show that \modelname outperforms state-of-the-art Chinese dialogue systems on both automatic and human evaluation, and demonstrates strong multi-task generalization on a variety of text understanding and generation tasks. In addition, we deploy \modelname to real-world applications such as Smart Speaker and Instant Message applications with fast inference. Our models and code will be made publicly available on ModelScope: https://modelscope.cn/models/damo/ChatPLUG-3.7B and Github: https://github.com/X-PLUG/ChatPLUG .

The development of large-scale Chinese language models is flourishing, yet there is a lack of corresponding capability assessments. Therefore, we propose a test to measure the multitask accuracy of large Chinese language models. This test encompasses four major domains, including medicine, law, psychology, and education, with 15 subtasks in medicine and 8 subtasks in education. We found that the best-performing models in the zero-shot setting outperformed the worst-performing models by nearly 18.6 percentage points on average. Across the four major domains, the highest average zero-shot accuracy of all models is 0.512. In the subdomains, only the GPT-3.5-turbo model achieved a zero-shot accuracy of 0.693 in clinical medicine, which was the highest accuracy among all models across all subtasks. All models performed poorly in the legal domain, with the highest zero-shot accuracy reaching only 0.239. By comprehensively evaluating the breadth and depth of knowledge across multiple disciplines, this test can more accurately identify the shortcomings of the models.

Large Language Models pre-trained with self-supervised learning have demonstrated impressive zero-shot generalization capabilities on a wide spectrum of tasks. In this work, we present WeLM: a well-read pre-trained language model for Chinese that is able to seamlessly perform different types of tasks with zero or few-shot demonstrations. WeLM is trained with 10B parameters by "reading" a curated high-quality corpus covering a wide range of topics. We show that WeLM is equipped with broad knowledge on various domains and languages. On 18 monolingual (Chinese) tasks, WeLM can significantly outperform existing pre-trained models with similar sizes and match the performance of models up to 25 times larger. WeLM also exhibits strong capabilities in multi-lingual and code-switching understanding, outperforming existing multilingual language models pre-trained on 30 languages. Furthermore, We collected human-written prompts for a large set of supervised datasets in Chinese and fine-tuned WeLM with multi-prompted training. The resulting model can attain strong generalization on unseen types of tasks and outperform the unsupervised WeLM in zero-shot learning. Finally, we demonstrate that WeLM has basic skills at explaining and calibrating the decisions from itself, which can be promising directions for future research. Our models can be applied from https://welm.weixin.qq.com/docs/api/.

Autonomous agents are able to draw on a wide variety of potential sources of task knowledge; however current approaches invariably focus on only one or two. Here we investigate the challenges and impact of exploiting diverse knowledge sources to learn online, in one-shot, new tasks for a simulated office mobile robot. The resulting agent, developed in the Soar cognitive architecture, uses the following sources of domain and task knowledge: interaction with the environment, task execution and search knowledge, human natural language instruction, and responses retrieved from a large language model (GPT-3). We explore the distinct contributions of these knowledge sources and evaluate the performance of different combinations in terms of learning correct task knowledge and human workload. Results show that an agent's online integration of diverse knowledge sources improves one-shot task learning overall, reducing human feedback needed for rapid and reliable task learning.

Knowledge graphs can represent information about the real-world using entities and their relations in a structured and semantically rich manner and they enable a variety of downstream applications such as question-answering, recommendation systems, semantic search, and advanced analytics. However, at the moment, building a knowledge graph involves a lot of manual effort and thus hinders their application in some situations and the automation of this process might benefit especially for small organizations. Automatically generating structured knowledge graphs from a large volume of natural language is still a challenging task and the research on sub-tasks such as named entity extraction, relation extraction, entity and relation linking, and knowledge graph construction aims to improve the state of the art of automatic construction and completion of knowledge graphs from text. The recent advancement of foundation models with billions of parameters trained in a self-supervised manner with large volumes of training data that can be adapted to a variety of downstream tasks has helped to demonstrate high performance on a large range of Natural Language Processing (NLP) tasks. In this context, one emerging paradigm is in-context learning where a language model is used as it is with a prompt that provides instructions and some examples to perform a task without changing the parameters of the model using traditional approaches such as fine-tuning. This way, no computing resources are needed for re-training/fine-tuning the models and the engineering effort is minimal. Thus, it would be beneficial to utilize such capabilities for generating knowledge graphs from text.

Conventional Transformer-based Video Question Answering (VideoQA) approaches generally encode frames independently through one or more image encoders followed by interaction between frames and question. However, such schema would incur significant memory use and inevitably slow down the training and inference speed. In this work, we present a highly efficient approach for VideoQA based on existing vision-language pre-trained models where we concatenate video frames to a $n\times n$ matrix and then convert it to one image. By doing so, we reduce the use of the image encoder from $n^{2}$ to $1$ while maintaining the temporal structure of the original video. Experimental results on MSRVTT and TrafficQA show that our proposed approach achieves state-of-the-art performance with nearly $4\times$ faster speed and only 30% memory use. We show that by integrating our approach into VideoQA systems we can achieve comparable, even superior, performance with a significant speed up for training and inference. We believe the proposed approach can facilitate VideoQA-related research by reducing the computational requirements for those who have limited access to budgets and resources. Our code will be made publicly available for research use.

Current methods for prompt learning in zeroshot scenarios widely rely on a development set with sufficient human-annotated data to select the best-performing prompt template a posteriori. This is not ideal because in a realworld zero-shot scenario of practical relevance, no labelled data is available. Thus, we propose a simple yet effective method for screening reasonable prompt templates in zero-shot text classification: Perplexity Selection (Perplection). We hypothesize that language discrepancy can be used to measure the efficacy of prompt templates, and thereby develop a substantiated perplexity-based scheme allowing for forecasting the performance of prompt templates in advance. Experiments show that our method leads to improved prediction performance in a realistic zero-shot setting, eliminating the need for any labelled examples.