Cross-encoders are effective passage re-rankers. But when re-ranking multiple passages at once, existing cross-encoders inefficiently optimize the output ranking over several input permutations, as their passage interactions are not permutation-invariant. Moreover, their high memory footprint constrains the number of passages during listwise training. To tackle these issues, we propose the Set-Encoder, a new cross-encoder architecture that (1) introduces inter-passage attention with parallel passage processing to ensure permutation invariance between input passages, and that (2) uses fused-attention kernels to enable training with more passages at a time. In experiments on TREC Deep Learning and TIREx, the Set-Encoder is more effective than previous cross-encoders with a similar number of parameters. Compared to larger models, the Set-Encoder is more efficient and either on par or even more effective.
The introduction of Vec2Text, a technique for inverting text embeddings, has raised serious privacy concerns within dense retrieval systems utilizing text embeddings, including those provided by OpenAI and Cohere. This threat comes from the ability for a malicious attacker with access to text embeddings to reconstruct the original text. In this paper, we investigate various aspects of embedding models that could influence the recoverability of text using Vec2Text. Our exploration involves factors such as distance metrics, pooling functions, bottleneck pre-training, training with noise addition, embedding quantization, and embedding dimensions -- aspects not previously addressed in the original Vec2Text paper. Through a thorough analysis of these factors, our aim is to gain a deeper understanding of the critical elements impacting the trade-offs between text recoverability and retrieval effectiveness in dense retrieval systems. This analysis provides valuable insights for practitioners involved in designing privacy-aware dense retrieval systems. Additionally, we propose a straightforward fix for embedding transformation that ensures equal ranking effectiveness while mitigating the risk of text recoverability. Furthermore, we extend the application of Vec2Text to the separate task of corpus poisoning, where, theoretically, Vec2Text presents a more potent threat compared to previous attack methods. Notably, Vec2Text does not require access to the dense retriever's model parameters and can efficiently generate numerous adversarial passages. In summary, this study highlights the potential threat posed by Vec2Text to existing dense retrieval systems, while also presenting effective methods to patch and strengthen such systems against such risks.
Federated search systems aggregate results from multiple search engines, selecting appropriate sources to enhance result quality and align with user intent. With the increasing uptake of Retrieval-Augmented Generation (RAG) pipelines, federated search can play a pivotal role in sourcing relevant information across heterogeneous data sources to generate informed responses. However, existing datasets, such as those developed in the past TREC FedWeb tracks, predate the RAG paradigm shift and lack representation of modern information retrieval challenges. To bridge this gap, we present FeB4RAG, a novel dataset specifically designed for federated search within RAG frameworks. This dataset, derived from 16 sub-collections of the widely used \beir benchmarking collection, includes 790 information requests (akin to conversational queries) tailored for chatbot applications, along with top results returned by each resource and associated LLM-derived relevance judgements. Additionally, to support the need for this collection, we demonstrate the impact on response generation of a high quality federated search system for RAG compared to a naive approach to federated search. We do so by comparing answers generated through the RAG pipeline through a qualitative side-by-side comparison. Our collection fosters and supports the development and evaluation of new federated search methods, especially in the context of RAG pipelines.
Text stemming is a natural language processing technique that is used to reduce words to their base form, also known as the root form. The use of stemming in IR has been shown to often improve the effectiveness of keyword-matching models such as BM25. However, traditional stemming methods, focusing solely on individual terms, overlook the richness of contextual information. Recognizing this gap, in this paper, we investigate the promising idea of using large language models (LLMs) to stem words by leveraging its capability of context understanding. With this respect, we identify three avenues, each characterised by different trade-offs in terms of computational cost, effectiveness and robustness : (1) use LLMs to stem the vocabulary for a collection, i.e., the set of unique words that appear in the collection (vocabulary stemming), (2) use LLMs to stem each document separately (contextual stemming), and (3) use LLMs to extract from each document entities that should not be stemmed, then use vocabulary stemming to stem the rest of the terms (entity-based contextual stemming). Through a series of empirical experiments, we compare the use of LLMs for stemming with that of traditional lexical stemmers such as Porter and Krovetz for English text. We find that while vocabulary stemming and contextual stemming fail to achieve higher effectiveness than traditional stemmers, entity-based contextual stemming can achieve a higher effectiveness than using Porter stemmer alone, under specific conditions.
This paper introduces a novel unsupervised technique that utilizes large language models (LLMs) to determine the most suitable dense retriever for a specific test(target) corpus. Selecting the appropriate dense retriever is vital for numerous IR applications that employ these retrievers, trained on public datasets, to encode or conduct searches within a new private target corpus. The effectiveness of a dense retriever can significantly diminish when applied to a target corpus that diverges in domain or task from the original training set. The problem becomes more pronounced in cases where the target corpus is unlabeled, e.g. in zero-shot scenarios, rendering direct evaluation of the model's effectiveness on the target corpus unattainable. Therefore, the unsupervised selection of an optimally pre-trained dense retriever, especially under conditions of domain shift, emerges as a critical challenge. Existing methodologies for ranking dense retrievers fall short in addressing these domain shift scenarios. To tackle this, our method capitalizes on LLMs to create pseudo-relevant queries, labels, and reference lists by analyzing a subset of documents from the target corpus. This allows for the ranking of dense retrievers based on their performance with these pseudo-relevant signals. Significantly, this strategy is the first to depend exclusively on the target corpus data, removing the necessity for training data and test labels. We assessed the effectiveness of our approach by compiling a comprehensive pool of cutting-edge dense retrievers and comparing our method against traditional dense retriever selection benchmarks. The findings reveal that our proposed solution surpasses the existing benchmarks in both the selection and ranking of dense retrievers.
Systematic reviews are crucial for evidence-based medicine as they comprehensively analyse published research findings on specific questions. Conducting such reviews is often resource- and time-intensive, especially in the screening phase, where abstracts of publications are assessed for inclusion in a review. This study investigates the effectiveness of using zero-shot large language models~(LLMs) for automatic screening. We evaluate the effectiveness of eight different LLMs and investigate a calibration technique that uses a predefined recall threshold to determine whether a publication should be included in a systematic review. Our comprehensive evaluation using five standard test collections shows that instruction fine-tuning plays an important role in screening, that calibration renders LLMs practical for achieving a targeted recall, and that combining both with an ensemble of zero-shot models saves significant screening time compared to state-of-the-art approaches.
Federated search, which involves integrating results from multiple independent search engines, will become increasingly pivotal in the context of Retrieval-Augmented Generation pipelines empowering LLM-based applications such as chatbots. These systems often distribute queries among various search engines, ranging from specialized (e.g., PubMed) to general (e.g., Google), based on the nature of user utterances. A critical aspect of federated search is resource selection - the selection of appropriate resources prior to issuing the query to ensure high-quality and rapid responses, and contain costs associated with calling the external search engines. However, current SOTA resource selection methodologies primarily rely on feature-based learning approaches. These methods often involve the labour intensive and expensive creation of training labels for each resource. In contrast, LLMs have exhibited strong effectiveness as zero-shot methods across NLP and IR tasks. We hypothesise that in the context of federated search LLMs can assess the relevance of resources without the need for extensive predefined labels or features. In this paper, we propose ReSLLM. Our ReSLLM method exploits LLMs to drive the selection of resources in federated search in a zero-shot setting. In addition, we devise an unsupervised fine tuning protocol, the Synthetic Label Augmentation Tuning (SLAT), where the relevance of previously logged queries and snippets from resources is predicted using an off-the-shelf LLM and then in turn used to fine-tune ReSLLM with respect to resource selection. Our empirical evaluation and analysis details the factors influencing the effectiveness of LLMs in this context. The results showcase the merits of ReSLLM for resource selection: not only competitive effectiveness in the zero-shot setting, but also obtaining large when fine-tuned using SLAT-protocol.
We describe team ielab from CSIRO and The University of Queensland's approach to the 2023 TREC Clinical Trials Track. Our approach was to use neural rankers but to utilise Large Language Models to overcome the issue of lack of training data for such rankers. Specifically, we employ ChatGPT to generate relevant patient descriptions for randomly selected clinical trials from the corpus. This synthetic dataset, combined with human-annotated training data from previous years, is used to train both dense and sparse retrievers based on PubmedBERT. Additionally, a cross-encoder re-ranker is integrated into the system. To further enhance the effectiveness of our approach, we prompting GPT-4 as a TREC annotator to provide judgments on our run files. These judgments are subsequently employed to re-rank the results. This architecture tightly integrates strong PubmedBERT-based rankers with the aid of SOTA Large Language Models, demonstrating a new approach to clinical trial retrieval.
In the field of information retrieval, Query Likelihood Models (QLMs) rank documents based on the probability of generating the query given the content of a document. Recently, advanced large language models (LLMs) have emerged as effective QLMs, showcasing promising ranking capabilities. This paper focuses on investigating the genuine zero-shot ranking effectiveness of recent LLMs, which are solely pre-trained on unstructured text data without supervised instruction fine-tuning. Our findings reveal the robust zero-shot ranking ability of such LLMs, highlighting that additional instruction fine-tuning may hinder effectiveness unless a question generation task is present in the fine-tuning dataset. Furthermore, we introduce a novel state-of-the-art ranking system that integrates LLM-based QLMs with a hybrid zero-shot retriever, demonstrating exceptional effectiveness in both zero-shot and few-shot scenarios. We make our codebase publicly available at https://github.com/ielab/llm-qlm.