Automatic medication mining from clinical and biomedical text has become a popular topic due to its real impact on healthcare applications and the recent development of powerful language models (LMs). However, fully-automatic extraction models still face obstacles to be overcome such that they can be deployed directly into clinical practice for better impacts. Such obstacles include their imbalanced performances on different entity types and clinical events. In this work, we examine current state-of-the-art pre-trained language models (PLMs) on such tasks, via fine-tuning including the monolingual model Med7 and multilingual large language model (LLM) XLM-RoBERTa. We compare their advantages and drawbacks using historical medication mining shared task data sets from n2c2-2018 challenges. We report the findings we get from these fine-tuning experiments such that they can facilitate future research on addressing them, for instance, how to combine their outputs, merge such models, or improve their overall accuracy by ensemble learning and data augmentation. MedMine is part of the M3 Initiative \url{https://github.com/HECTA-UoM/M3}

Existing large language models have to run K times to generate a sequence of K tokens. Our approach relies on the observation that adjacent tokens in a sequence usually have strong correlations and the next token in a sequence can be reasonably guessed or inferred based on the preceding ones. Experiments and analysis demonstrate the effectiveness of our approach in lowering inference latency, achieving up to 1.4x speedup while preserving high performance. Large language models (LLMs) (Brown et al., 2020; OpenAI, 2023; Touvron et al., 2023; Chowdhery et al., 2022; Biderman et al., 2023; Smith et al., 2022) have revolutionized the field of natural language generation for their abilities in generating satisfactory text across various application domains. The excellent performance benefits greatly from the scaling of model size (100B+ parameters), but at the same time, the fact remains that a single decoding step gets slower as the model gets larger. In addition to the immense computation introduced by larger models, a larger memory footprint is also a major factor causing slower inference of LLMs (Dao et al., 2022; Pope et al., 2023). This large memory footprint includes the trained model parameters, the temporary state used during inference, and in addition to these, the KV cache is also stored in memory.

Contrastive pretrained large Vision-Language Models (VLMs) like CLIP have revolutionized visual representation learning by providing good performance on downstream datasets. VLMs are 0-shot adapted to a downstream dataset by designing prompts that are relevant to the dataset. Such prompt engineering makes use of domain expertise and a validation dataset. Meanwhile, recent developments in generative pretrained models like GPT-4 mean they can be used as advanced internet search tools. They can also be manipulated to provide visual information in any structure. In this work, we show that GPT-4 can be used to generate text that is visually descriptive and how this can be used to adapt CLIP to downstream tasks. We show considerable improvements in 0-shot transfer accuracy on specialized fine-grained datasets like EuroSAT (~7%), DTD (~7%), SUN397 (~4.6%), and CUB (~3.3%) when compared to CLIP's default prompt. We also design a simple few-shot adapter that learns to choose the best possible sentences to construct generalizable classifiers that outperform the recently proposed CoCoOP by ~2% on average and by over 4% on 4 specialized fine-grained datasets. The code, prompts, and auxiliary text dataset is available at https://github.com/mayug/VDT-Adapter.

The derivation of mathematical results in specialised fields, using Large Language Models (LLMs), is an emerging research direction that can help identify models' limitations, and potentially support mathematical discovery. In this paper, we leverage a symbolic engine to generate derivations of equations at scale, and investigate the capabilities of LLMs when deriving goal equations from premises. Specifically, we employ in-context learning for GPT and fine-tune a range of T5 models to compare the robustness and generalisation of pre-training strategies to specialised models. Empirical results show that fine-tuned FLAN-T5-large (MathT5) outperforms GPT models on all static and out-of-distribution test sets in conventional scores. However, an in-depth analysis reveals that the fine-tuned models are more sensitive to perturbations involving unseen symbols and (to a lesser extent) changes to equation structure. In addition, we analyse 1.7K equations, and over 200 derivations, to highlight common reasoning errors such as the inclusion of incorrect, irrelevant, and redundant equations. Finally, we explore the suitability of existing metrics for evaluating mathematical derivations and find evidence that, while they can capture general properties such as sensitivity to perturbations, they fail to highlight fine-grained reasoning errors and essential differences between models. Overall, this work demonstrates that training models on synthetic data may improve their math capabilities beyond much larger LLMs, but current metrics are not appropriately assessing the quality of generated mathematical text.

Logical reasoning consistently plays a fundamental and significant role in the domains of knowledge engineering and artificial intelligence. Recently, Large Language Models (LLMs) have emerged as a noteworthy innovation in natural language processing (NLP), exhibiting impressive achievements across various classic NLP tasks. However, the question of whether LLMs can effectively address the task of logical reasoning, which requires gradual cognitive inference similar to human intelligence, remains unanswered. To this end, we aim to bridge this gap and provide comprehensive evaluations in this paper. Firstly, to offer systematic evaluations, we select fifteen typical logical reasoning datasets and organize them into deductive, inductive, abductive and mixed-form reasoning settings. Considering the comprehensiveness of evaluations, we include three representative LLMs (i.e., text-davinci-003, ChatGPT and BARD) and evaluate them on all selected datasets under zero-shot, one-shot and three-shot settings. Secondly, different from previous evaluations relying only on simple metrics (e.g., accuracy), we propose fine-level evaluations from objective and subjective manners, covering both answers and explanations. Additionally, to uncover the logical flaws of LLMs, problematic cases will be attributed to five error types from two dimensions, i.e., evidence selection process and reasoning process. Thirdly, to avoid the influences of knowledge bias and purely focus on benchmarking the logical reasoning capability of LLMs, we propose a new dataset with neutral content. It contains 3,000 samples and covers deductive, inductive and abductive settings. Based on the in-depth evaluations, this paper finally forms a general evaluation scheme of logical reasoning capability from six dimensions. It reflects the pros and cons of LLMs and gives guiding directions for future works.

While large text-to-image models are able to synthesize "novel" images, these images are necessarily a reflection of the training data. The problem of data attribution in such models -- which of the images in the training set are most responsible for the appearance of a given generated image -- is a difficult yet important one. As an initial step toward this problem, we evaluate attribution through "customization" methods, which tune an existing large-scale model toward a given exemplar object or style. Our key insight is that this allows us to efficiently create synthetic images that are computationally influenced by the exemplar by construction. With our new dataset of such exemplar-influenced images, we are able to evaluate various data attribution algorithms and different possible feature spaces. Furthermore, by training on our dataset, we can tune standard models, such as DINO, CLIP, and ViT, toward the attribution problem. Even though the procedure is tuned towards small exemplar sets, we show generalization to larger sets. Finally, by taking into account the inherent uncertainty of the problem, we can assign soft attribution scores over a set of training images.

Large language models~(LLMs) are instruction followers, but it can be challenging to find the best instruction for different situations, especially for black-box LLMs on which backpropagation is forbidden. Instead of directly optimizing the discrete instruction, we optimize a low-dimensional soft prompt applied to an open-source LLM to generate the instruction for the black-box LLM. On each iteration of the proposed method, which we call InstructZero, a soft prompt is converted into an instruction using the open-source LLM, which is then submitted to the black-box LLM for zero-shot evaluation, and the performance is sent to Bayesian optimization to produce new soft prompts improving the zero-shot performance. We evaluate InstructZero on different combinations of open-source LLMs and APIs including Vicuna and ChatGPT. Our results show that InstructZero outperforms SOTA auto-instruction methods across a variety of downstream tasks. Our code and data are publicly available at https://github.com/Lichang-Chen/InstructZero.

Zero-shot quantization is a promising approach for developing lightweight deep neural networks when data is inaccessible owing to various reasons, including cost and issues related to privacy. By exploiting the learned parameters ($\mu$ and $\sigma$) of batch normalization layers in an FP32-pre-trained model, zero-shot quantization schemes focus on generating synthetic data. Subsequently, they distill knowledge from the pre-trained model (teacher) to the quantized model (student) such that the quantized model can be optimized with the synthetic dataset. However, thus far, zero-shot quantization has primarily been discussed in the context of quantization-aware training methods, which require task-specific losses and long-term optimization as much as retraining. We thus introduce a post-training quantization scheme for zero-shot quantization that produces high-quality quantized networks within a few hours. Furthermore, we propose a framework called Genie~that generates data suited for quantization. With the data synthesized by Genie, we can produce robust quantized models without real datasets, which is comparable to few-shot quantization. We also propose a post-training quantization algorithm to enhance the performance of quantized models. By combining them, we can bridge the gap between zero-shot and few-shot quantization while significantly improving the quantization performance compared to that of existing approaches. In other words, we can obtain a unique state-of-the-art zero-shot quantization approach. The code is available at \url{https://github.com/SamsungLabs/Genie}.

For most people, the idea of using artificial intelligence tools in daily life--or even just messing around with them--has only become mainstream in recent months, with new releases of generative AI tools from a slew of big tech companies and startups, like OpenAI's ChatGPT and Google's Bard. But behind the scenes, the technology has been proliferating for years, along with questions about how best to evaluate and secure these new AI systems. On Monday, Microsoft is revealing details about the team within the company that since 2018 has been tasked with figuring out how to attack AI platforms to reveal their weaknesses. In the five years since its formation, Microsoft's AI red team has grown from what was essentially an experiment into a full interdisciplinary team of machine learning experts, cybersecurity researchers, and even social engineers. The group works to communicate its findings within Microsoft and across the tech industry using the traditional parlance of digital security, so the ideas will be accessible rather than requiring specialized AI knowledge that many people and organizations don't yet have.

The researchers asked language models where they stand on various topics, such as feminism and democracy. They used the answers to plot them on a graph known as a political compass, and then tested whether retraining models on even more politically biased training data changed their behavior and ability to detect hate speech and misinformation (it did). The research is described in a peer-reviewed paper that won the best paper award at the Association for Computational Linguistics conference last month. As AI language models are rolled out into products and services used by millions of people, understanding their underlying political assumptions and biases could not be more important. That's because they have the potential to cause real harm.