It is a common practice in natural language processing to pre-train a single model on a general domain and then fine-tune it for downstream tasks. However, when it comes to Large Language Models, fine-tuning the entire model can be computationally expensive, resulting in very intensive energy consumption. As a result, several Parameter Efficient Fine-Tuning (PEFT) approaches were recently proposed. One of the most popular approaches is low-rank adaptation (LoRA), where the key insight is decomposing the update weights of the pre-trained model into two low-rank matrices. However, the proposed approaches either use the same rank value across all different weight matrices, which has been shown to be a sub-optimal choice, or do not use any quantization technique, one of the most important factors when it comes to a model's energy consumption. In this work, we propose Bayesian-LoRA which approaches low-rank adaptation and quantization from a Bayesian perspective by employing a prior distribution on both quantization levels and rank values. As a result, B-LoRA is able to fine-tune a pre-trained model on a specific downstream task, finding the optimal rank values and quantization levels for every low-rank matrix. We validate the proposed model by fine-tuning a pre-trained DeBERTaV3 on the GLUE benchmark. Moreover, we compare it to relevant baselines and present both qualitative and quantitative results, showing how the proposed approach is able to learn optimal-rank quantized matrices. B-LoRA performs on par with or better than the baselines while reducing the total number of bit operations by roughly 70% compared to the baseline methods.

Large language models (LLMs) have revolutionized Natural Language Processing (NLP) by minimizing the need for complex feature engineering. However, the application of LLMs in specialized domains like biopharmaceuticals and chemistry remains largely unexplored. These fields are characterized by intricate terminologies, specialized knowledge, and a high demand for precision areas where general purpose LLMs often fall short. In this study, we introduce PharmaGPT, a suite of domain specilized LLMs with 13 billion and 70 billion parameters, specifically trained on a comprehensive corpus tailored to the Bio-Pharmaceutical and Chemical domains. Our evaluation shows that PharmaGPT surpasses existing general models on specific-domain benchmarks such as NAPLEX, demonstrating its exceptional capability in domain-specific tasks. Remarkably, this performance is achieved with a model that has only a fraction, sometimes just one-tenth-of the parameters of general-purpose large models. This advancement establishes a new benchmark for LLMs in the bio-pharmaceutical and chemical fields, addressing the existing gap in specialized language modeling. It also suggests a promising path for enhanced research and development, paving the way for more precise and effective NLP applications in these areas.

A rapidly growing number of applications rely on a small set of closed-source language models (LMs). This dependency might introduce novel security risks if LMs develop self-recognition capabilities. Inspired by human identity verification methods, we propose a novel approach for assessing self-recognition in LMs using model-generated "security questions". Our test can be externally administered to keep track of frontier models as it does not require access to internal model parameters or output probabilities. We use our test to examine self-recognition in ten of the most capable open- and closed-source LMs currently publicly available. Our extensive experiments found no empirical evidence of general or consistent self-recognition in any examined LM. Instead, our results suggest that given a set of alternatives, LMs seek to pick the "best" answer, regardless of its origin. Moreover, we find indications that preferences about which models produce the best answers are consistent across LMs. We additionally uncover novel insights on position bias considerations for LMs in multiple-choice settings.

We explore using Large Language Models (LLMs) to generate application code that automates health insurance processes from text-based policies. We target blockchain-based smart contracts as they offer immutability, verifiability, scalability, and a trustless setting: any number of parties can use the smart contracts, and they need not have previously established trust relationships with each other. Our methodology generates outputs at increasing levels of technical detail: (1) textual summaries, (2) declarative decision logic, and (3) smart contract code with unit tests. We ascertain LLMs are good at the task (1), and the structured output is useful to validate tasks (2) and (3). Declarative languages (task 2) are often used to formalize healthcare policies, but their execution on blockchain is non-trivial. Hence, task (3) attempts to directly automate the process using smart contracts. To assess the LLM output, we propose completeness, soundness, clarity, syntax, and functioning code as metrics. Our evaluation employs three health insurance policies (scenarios) with increasing difficulty from Medicare's official booklet. Our evaluation uses GPT-3.5 Turbo, GPT-3.5 Turbo 16K, GPT-4, GPT-4 Turbo and CodeLLaMA. Our findings confirm that LLMs perform quite well in generating textual summaries. Although outputs from tasks (2)-(3) are useful starting points, they require human oversight: in multiple cases, even "runnable" code will not yield sound results; the popularity of the target language affects the output quality; and more complex scenarios still seem a bridge too far. Nevertheless, our experiments demonstrate the promise of LLMs for translating textual process descriptions into smart contracts.

Our team participated in the BioASQ 2024 Task12b and Synergy tasks to build a system that can answer biomedical questions by retrieving relevant articles and snippets from the PubMed database and generating exact and ideal answers. We propose a two-level information retrieval and question-answering system based on pre-trained large language models (LLM), focused on LLM prompt engineering and response post-processing. We construct prompts with in-context few-shot examples and utilize post-processing techniques like resampling and malformed response detection. We compare the performance of various pre-trained LLM models on this challenge, including Mixtral, OpenAI GPT and Llama2. Our best-performing system achieved 0.14 MAP score on document retrieval, 0.05 MAP score on snippet retrieval, 0.96 F1 score for yes/no questions, 0.38 MRR score for factoid questions and 0.50 F1 score for list questions in Task 12b.

The task of writing rap is challenging and involves producing complex rhyming schemes, yet meaningful lyrics. In this work, we propose Raply, a fine-tuned GPT-2 model capable of producing meaningful rhyming text in the style of rap. In addition to its rhyming capabilities, the model is able to generate less offensive content. It was achieved through the fine-tuning the model on a new dataset Mitislurs, a profanity-mitigated corpus. We evaluate the output of the model on two criteria: 1) rhyming based on the rhyme density metric; 2) profanity content, using the list of profanities for the English language. To our knowledge, this is the first attempt at profanity mitigation for rap lyrics generation.

Is it always necessary to compute tokens from shallow to deep layers in Transformers? The continued success of vanilla Transformers and their variants suggests an undoubted "yes". In this work, however, we attempt to break the depth-ordered convention by proposing a novel architecture dubbed mixture-of-modules (MoM), which is motivated by an intuition that any layer, regardless of its position, can be used to compute a token as long as it possesses the needed processing capabilities. The construction of MoM starts from a finite set of modules defined by multi-head attention and feed-forward networks, each distinguished by its unique parameterization. Two routers then iteratively select attention modules and feed-forward modules from the set to process a token. The selection dynamically expands the computation graph in the forward pass of the token, culminating in an assembly of modules. We show that MoM provides not only a unified framework for Transformers and their numerous variants but also a flexible and learnable approach for reducing redundancy in Transformer parameterization. We pre-train various MoMs using OpenWebText. Empirical results demonstrate that MoMs, of different parameter counts, consistently outperform vanilla transformers on both GLUE and XSUM benchmarks. More interestingly, with a fixed parameter budget, MoM-large enables an over 38% increase in depth for computation graphs compared to GPT-2-large, resulting in absolute gains of 1.4 on GLUE and 1 on XSUM. On the other hand, MoM-large also enables an over 60% reduction in depth while involving more modules per layer, yielding a 16% reduction in TFLOPs and a 43% decrease in memory usage compared to GPT-2-large, while maintaining comparable performance.

Flowcharts are graphical tools for representing complex concepts in concise visual representations. This paper introduces the FlowLearn dataset, a resource tailored to enhance the understanding of flowcharts. FlowLearn contains complex scientific flowcharts and simulated flowcharts. The scientific subset contains 3,858 flowcharts sourced from scientific literature and the simulated subset contains 10,000 flowcharts created using a customizable script. The dataset is enriched with annotations for visual components, OCR, Mermaid code representation, and VQA question-answer pairs. Despite the proven capabilities of Large Vision-Language Models (LVLMs) in various visual understanding tasks, their effectiveness in decoding flowcharts - a crucial element of scientific communication - has yet to be thoroughly investigated. The FlowLearn test set is crafted to assess the performance of LVLMs in flowchart comprehension. Our study thoroughly evaluates state-of-the-art LVLMs, identifying existing limitations and establishing a foundation for future enhancements in this relatively underexplored domain. For instance, in tasks involving simulated flowcharts, GPT-4V achieved the highest accuracy (58%) in counting the number of nodes, while Claude recorded the highest accuracy (83%) in OCR tasks. Notably, no single model excels in all tasks within the FlowLearn framework, highlighting significant opportunities for further development.

Currently training a large event-text model still poses a huge challenge due to the shortage of paired event-text data. In response to this challenge, CEIA learns to align event and image data as an alternative instead of directly aligning event and text data. Specifically, we leverage the rich event-image datasets to learn an event embedding space aligned with the image space of CLIP through contrastive learning. In this way, event and text data are naturally aligned via using image data as a bridge. Particularly, CEIA offers two distinct advantages. First, it allows us to take full advantage of the existing event-image datasets to make up the shortage of large-scale event-text datasets. Second, leveraging more training data, it also exhibits the flexibility to boost performance, ensuring scalable capability. In highlighting the versatility of our framework, we make extensive evaluations through a diverse range of event-based multi-modal applications, such as object recognition, event-image retrieval, event-text retrieval, and domain adaptation. The outcomes demonstrate CEIA's distinct zero-shot superiority over existing methods on these applications.

We introduce SciQAG, a novel framework for automatically generating high-quality science question-answer pairs from a large corpus of scientific literature based on large language models (LLMs). SciQAG consists of a QA generator and a QA evaluator, which work together to extract diverse and research-level questions and answers from scientific papers. Utilizing this framework, we construct a large-scale, high-quality, open-ended science QA dataset containing 188,042 QA pairs extracted from 22,743 scientific papers across 24 scientific domains. We also introduce SciQAG-24D, a new benchmark task designed to evaluate the science question-answering ability of LLMs. Extensive experiments demonstrate that fine-tuning LLMs on the SciQAG dataset significantly improves their performance on both open-ended question answering and scientific tasks. To foster research and collaboration, we make the datasets, models, and evaluation codes publicly available, contributing to the advancement of science question answering and developing more interpretable and reasoning-capable AI systems.