Pretrained language models (PLMs) are trained on massive corpora, but often need to specialize to specific domains. A parameter-efficient adaptation method suggests training an adapter for each domain on the task of language modeling. This leads to good in-domain scores but can be impractical for domain- or resource-restricted settings. A solution is to use a related-domain adapter for the novel domain at test time. In this paper, we introduce AdapterSoup, an approach that performs weight-space averaging of adapters trained on different domains. Our approach is embarrassingly parallel: first, we train a set of domain-specific adapters; then, for each novel domain, we determine which adapters should be averaged at test time. We present extensive experiments showing that AdapterSoup consistently improves performance to new domains without extra training. We also explore weight averaging of adapters trained on the same domain with different hyper-parameters, and show that it preserves the performance of a PLM on new domains while obtaining strong in-domain results. We explore various approaches for choosing which adapters to combine, such as text clustering and semantic similarity. We find that using clustering leads to the most competitive results on novel domains.

Task-oriented dialog systems empower users to accomplish their goals by facilitating intuitive and expressive natural language interactions. State-of-the-art approaches in task-oriented dialog systems formulate the problem as a conditional sequence generation task and fine-tune pre-trained causal language models in the supervised setting. This requires labeled training data for each new domain or task, and acquiring such data is prohibitively laborious and expensive, thus making it a bottleneck for scaling systems to a wide range of domains. To overcome this challenge, we introduce a novel Zero-Shot generalizable end-to-end Task-oriented Dialog system, ZS-ToD, that leverages domain schemas to allow for robust generalization to unseen domains and exploits effective summarization of the dialog history. We employ GPT-2 as a backbone model and introduce a two-step training process where the goal of the first step is to learn the general structure of the dialog data and the second step optimizes the response generation as well as intermediate outputs, such as dialog state and system actions. As opposed to state-of-the-art systems that are trained to fulfill certain intents in the given domains and memorize task-specific conversational patterns, ZS-ToD learns generic task-completion skills by comprehending domain semantics via domain schemas and generalizing to unseen domains seamlessly. We conduct an extensive experimental evaluation on SGD and SGD-X datasets that span up to 20 unique domains and ZS-ToD outperforms state-of-the-art systems on key metrics, with an improvement of +17% on joint goal accuracy and +5 on inform. Additionally, we present a detailed ablation study to demonstrate the effectiveness of the proposed components and training mechanism

The large language model called ChatGPT has drawn extensively attention because of its human-like expression and reasoning abilities. In this study, we investigate the feasibility of using ChatGPT in experiments on using ChatGPT to translate radiology reports into plain language for patients and healthcare providers so that they are educated for improved healthcare. Radiology reports from 62 low-dose chest CT lung cancer screening scans and 76 brain MRI metastases screening scans were collected in the first half of February for this study. According to the evaluation by radiologists, ChatGPT can successfully translate radiology reports into plain language with an average score of 4.27 in the five-point system with 0.08 places of information missing and 0.07 places of misinformation. In terms of the suggestions provided by ChatGPT, they are general relevant such as keeping following-up with doctors and closely monitoring any symptoms, and for about 37% of 138 cases in total ChatGPT offers specific suggestions based on findings in the report. ChatGPT also presents some randomness in its responses with occasionally over-simplified or neglected information, which can be mitigated using a more detailed prompt. Furthermore, ChatGPT results are compared with a newly released large model GPT-4, showing that GPT-4 can significantly improve the quality of translated reports. Our results show that it is feasible to utilize large language models in clinical education, and further efforts are needed to address limitations and maximize their potential.

Intelligent writing assistants powered by large language models (LLMs) are more popular today than ever before, but their further widespread adoption is precluded by sub-optimal performance. In this position paper, we argue that a major reason for this sub-optimal performance and adoption is a singular focus on the information content of language while ignoring its social aspects. We analyze the different dimensions of these social factors in the context of writing assistants and propose their incorporation into building smarter, more effective, and truly personalized writing assistants that would enrich the user experience and contribute to increased user adoption.

We present new benchmarks on evaluation code generation models: MBXP and Multilingual HumanEval, and MathQA-X. These datasets cover over 10 programming languages and are generated using a scalable conversion framework that transpiles prompts and test cases from the original Python datasets into the corresponding data in the target language. Using these benchmarks, we are able to assess the performance of code generation models in a multi-lingual fashion, and discovered generalization ability of language models on out-of-domain languages, advantages of multi-lingual models over mono-lingual, the ability of few-shot prompting to teach the model new languages, and zero-shot translation abilities even on mono-lingual settings. Furthermore, we use our code generation model to perform large-scale bootstrapping to obtain synthetic canonical solutions in several languages, which can be used for other code-related evaluations such as code insertion, robustness, or summarization tasks. Overall, our benchmarks represents a significant step towards a deeper understanding of language models' code generation abilities. We publicly release our code and datasets at https://github.com/amazon-research/mxeval.

The potential for pre-trained large language models (LLMs) to use natural language feedback at inference time has been an exciting recent development. We build upon this observation by formalizing an algorithm for learning from natural language feedback at training time instead, which we call Imitation learning from Language Feedback (ILF). ILF requires only a small amount of human-written feedback during training and does not require the same feedback at test time, making it both user-friendly and sample-efficient. We further show that ILF can be seen as a form of minimizing the KL divergence to the ground truth distribution and demonstrate a proof-of-concept on a neural program synthesis task. We use ILF to improve a Codegen-Mono 6.1B model's pass@1 rate by 38% relative (and 10% absolute) on the Mostly Basic Python Problems (MBPP) benchmark, outperforming both fine-tuning on MBPP and fine-tuning on repaired programs written by humans. Overall, our results suggest that learning from human-written natural language feedback is both more effective and sample-efficient than training exclusively on demonstrations for improving an LLM's performance on code generation tasks.

Emerging events, such as the COVID pandemic and the Ukraine Crisis, require a time-sensitive comprehensive understanding of the situation to allow for appropriate decision-making and effective action response. Automated generation of situation reports can significantly reduce the time, effort, and cost for domain experts when preparing their official human-curated reports. However, AI research toward this goal has been very limited, and no successful trials have yet been conducted to automate such report generation. We propose SmartBook, a novel task formulation targeting situation report generation, which consumes large volumes of news data to produce a structured situation report with multiple hypotheses (claims) summarized and grounded with rich links to factual evidence. We realize SmartBook for the Ukraine-Russia crisis by automatically generating intelligence analysis reports to assist expert analysts. The machine-generated reports are structured in the form of timelines, with each timeline organized by major events (or chapters), corresponding strategic questions (or sections) and their grounded summaries (or section content). Our proposed framework automatically detects real-time event-related strategic questions, which are more directed than manually-crafted analyst questions, which tend to be too complex, hard to parse, vague and high-level. Results from thorough qualitative evaluations show that roughly 82% of the questions in Smartbook have strategic importance, with at least 93% of the sections in the report being tactically useful. Further, experiments show that expert analysts tend to add more information into the SmartBook reports, with only 2.3% of the existing tokens being deleted, meaning SmartBook can serve as a useful foundation for analysts to build upon when creating intelligence reports.

The Object Constraint Language (OCL) is a declarative language that adds constraints and object query expressions to MOF models. Despite its potential to provide precision and conciseness to UML models, the unfamiliar syntax of OCL has hindered its adoption. Recent advancements in LLMs, such as GPT-3, have shown their capability in many NLP tasks, including semantic parsing and text generation. Codex, a GPT-3 descendant, has been fine-tuned on publicly available code from GitHub and can generate code in many programming languages. We investigate the reliability of OCL constraints generated by Codex from natural language specifications. To achieve this, we compiled a dataset of 15 UML models and 168 specifications and crafted a prompt template with slots to populate with UML information and the target task, using both zero- and few-shot learning methods. By measuring the syntactic validity and execution accuracy metrics of the generated OCL constraints, we found that enriching the prompts with UML information and enabling few-shot learning increases the reliability of the generated OCL constraints. Furthermore, the results reveal a close similarity based on sentence embedding between the generated OCL constraints and the human-written ones in the ground truth, implying a level of clarity and understandability in the generated OCL constraints by Codex.

Artificial Intelligence (AI) has made incredible progress recently. On the one hand, advanced foundation models like ChatGPT can offer powerful conversation, in-context learning and code generation abilities on a broad range of open-domain tasks. They can also generate high-level solution outlines for domain-specific tasks based on the common sense knowledge they have acquired. However, they still face difficulties with some specialized tasks because they lack enough domain-specific data during pre-training or they often have errors in their neural network computations on those tasks that need accurate executions. On the other hand, there are also many existing models and systems (symbolic-based or neural-based) that can do some domain-specific tasks very well. However, due to the different implementation or working mechanisms, they are not easily accessible or compatible with foundation models. Therefore, there is a clear and pressing need for a mechanism that can leverage foundation models to propose task solution outlines and then automatically match some of the sub-tasks in the outlines to the off-the-shelf models and systems with special functionalities to complete them. Inspired by this, we introduce TaskMatrix.AI as a new AI ecosystem that connects foundation models with millions of APIs for task completion. Unlike most previous work that aimed to improve a single AI model, TaskMatrix.AI focuses more on using existing foundation models (as a brain-like central system) and APIs of other AI models and systems (as sub-task solvers) to achieve diversified tasks in both digital and physical domains. As a position paper, we will present our vision of how to build such an ecosystem, explain each key component, and use study cases to illustrate both the feasibility of this vision and the main challenges we need to address next.

ChatGPT has enabled access to AI-generated writing for the masses, and within just a few months, this product has disrupted the knowledge economy, initiating a culture shift in the way people work, learn, and write. The need to discriminate human writing from AI is now both critical and urgent, particularly in domains like higher education and academic writing, where AI had not been a significant threat or contributor to authorship. Addressing this need, we developed a method for discriminating text generated by ChatGPT from (human) academic scientists, relying on prevalent and accessible supervised classification methods. We focused on how a particular group of humans, academic scientists, write differently than ChatGPT, and this targeted approach led to the discovery of new features for discriminating (these) humans from AI; as examples, scientists write long paragraphs and have a penchant for equivocal language, frequently using words like but, however, and although. With a set of 20 features, including the aforementioned ones and others, we built a model that assigned the author, as human or AI, at well over 99% accuracy, resulting in 20 times fewer misclassified documents compared to the field-leading approach. This strategy for discriminating a particular set of humans writing from AI could be further adapted and developed by others with basic skills in supervised classification, enabling access to many highly accurate and targeted models for detecting AI usage in academic writing and beyond.