While it is widely agreed that large language models (LLMs) store concepts from multiple semantic hierarchies, much remains unknown regarding the structure of this storage. The correspondence between the functional roles of LLM components and the semantic hierarchies of knowledge remains underexplored in the current literature. For example, is information organized hierarchically within sections of an LLM? We take an initial step towards causally examining the correspondence between hierarchical concepts and the multi-granular structures (layers and attention heads) of various models. Specifically, we generate a dataset of semantic hierarchies and investigate their storage locations in six LLMs using activation patching, a causal intervention technique. At the layer level, our findings show a moderate indication that concepts at finer levels of granularity are stored around 61-78% of the time (p < 0.01) before those at coarser granularity. There is evidence for this trend at the attention level; however, the high variability in attention level results suggests that concepts are stored across attention heads rather than within. Our results offer insight into semantic organization within LLMs.

Health literacy plays a critical role in ensuring people can access, understand, and act on medical information. However, much of the health content available today is too complex for many people, and simplifying these texts manually is time-consuming and difficult to do at scale.To overcome this, we developed a new framework for automatically generating health answers at multiple, precisely controlled complexity levels.We began with a thorough analysis of 166 linguistic features, which we then refined into 13 key metrics that reliably differentiate between simple and complex medical texts. From these metrics, we derived a robust complexity scoring formula, combining them with weights learned from a logistic regression model. This formula allowed us to create a large, multi-level dataset of health question-answer pairs covering 21 distinct complexity levels, ranging from elementary patient-friendly explanations to highly technical summaries.Finally, we fine-tuned a Llama-3.1-8B-Instruct model using “control codes” on this dataset, giving users precise control over the complexity of the generated text and empowering them to select the level of detail and technicality they need.

The broad integration of neural retrieval models into Information Retrieval (IR) systems is significantly impeded by the high cost and laborious process associated with the manual labelling of training data. Similarly, synthetic training data generation, a potential workaround, often requires expensive computational resources due to the reliance on large language models. This work explored the potential of small language models for efficiently creating high-quality synthetic datasets to train neural retrieval models. We aim to identify an optimal method to generate synthetic datasets, enabling training neural reranking models in document collections where annotated data is unavailable. We introduce a novel methodology, grounded in the principles of information theory, to select the most appropriate documents to be used as context for question generation. Then, we employ a small language model for zero-shot conditional question generation, supplemented by a filtering mechanism to ensure the quality of generated questions. Extensive evaluation on five datasets unveils the potential of our approach, outperforming unsupervised retrieval methods such as BM25 and pretrained monoT5. Our findings indicate that an efficiently generated “silver-standard” dataset allows effective training of neural rerankers in unlabeled scenarios. To ensure reproducibility and facilitate wider application, we will release a code repository featuring an accessible API for zero-shot synthetic question generation.

Transformer-based “behemoths” have grown in popularity, as well as structurally, shattering multiple NLP benchmarks along the way. However, their real-world usability remains a question. In this work, we empirically assess the feasibility of applying transformer-based models in real-world ad-hoc retrieval applications by comparison to a “greener and more sustainable” alternative, comprising only 620 trainable parameters. We present an analysis of their efficacy and efficiency and show that considering limited computational resources, the lighter model running on the CPU achieves a 3 to 20 times speedup in training and 7 to 47 times in inference while maintaining a comparable retrieval performance. Code to reproduce the efficiency experiments is available on “https://github.com/bioinformatics-ua/EACL2021-reproducibility/“.

The Covid-19 pandemic urged the scientific community to join efforts at an unprecedented scale, leading to faster than ever dissemination of data and results, which in turn motivated more research works. This paper presents and discusses information retrieval models aimed at addressing the challenge of searching the large number of publications that stem from these studies. The model presented, based on classical baselines followed by an interaction based neural ranking model, was evaluated and evolved within the TREC Covid challenge setting. Results on this dataset show that, when starting with a strong baseline, our light neural ranking model can achieve results that are comparable to other model architectures that use very large number of parameters.