Retrieval-Augmented Generation (RAG) has recently emerged as a method to extend beyond the pre-trained knowledge of Large Language Models by augmenting the original prompt with relevant passages or documents retrieved by an Information Retrieval (IR) system. RAG has become increasingly important for Generative AI solutions, especially in enterprise settings or in any domain in which knowledge is constantly refreshed and cannot be memorized in the LLM. We argue here that the retrieval component of RAG systems, be it dense or sparse, deserves increased attention from the research community, and accordingly, we conduct the first comprehensive and systematic examination of the retrieval strategy of RAG systems. We focus, in particular, on the type of passages IR systems within a RAG solution should retrieve. Our analysis considers multiple factors, such as the relevance of the passages included in the prompt context, their position, and their number. One counter-intuitive finding of this work is that the retriever's highest-scoring documents that are not directly relevant to the query (e.g., do not contain the answer) negatively impact the effectiveness of the LLM. Even more surprising, we discovered that adding random documents in the prompt improves the LLM accuracy by up to 35%. These results highlight the need to investigate the appropriate strategies when integrating retrieval with LLMs, thereby laying the groundwork for future research in this area.
12 December 2016 in ACM Transactions on Information Systems
Learning-to-Rank models based on additive ensembles of regression trees have been proven to be very effective for scoring query results returned by large-scale Web search engines. Unfortunately, the computational cost of scoring thousands of candidate documents by traversing large ensembles of trees is high. Thus, several works have investigated solutions aimed at improving the efficiency of document scoring by exploiting advanced features of modern CPUs and memory hierarchies. In this article, we present Q uick S corer , a new algorithm that adopts a novel cache-efficient representation of a given tree ensemble, performs an interleaved traversal by means of fast bitwise operations, and supports ensembles of oblivious trees. An extensive and detailed test assessment is conducted on two standard Learning-to-Rank datasets and on a novel very large dataset we made publicly available for conducting significant efficiency tests. The experiments show unprecedented speedups over the best state-of-the-art baselines ranging from 1.9 × to 6.6 × . The analysis of low-level profiling traces shows that Q uick S corer efficiency is due to its cache-aware approach in terms of both data layout and access patterns and to a control flow that entails very low branch mis-prediction rates.
Pseudo-relevance feedback mechanisms, from Rocchio to the relevance models, have shown the usefulness of expanding and reweighting the users’ initial queries using information occurring in an initial set of retrieved documents, known as the pseudo-relevant set. Recently, dense retrieval – through the use of neural contextual language models such as BERT for analysing the documents’ and queries’ contents and computing their relevance scores – has shown a promising performance on several information retrieval tasks still relying on the traditional inverted index for identifying documents relevant to a query. Two different dense retrieval families have emerged: the use of single embedded representations for each passage and query, e.g., using BERT’s [CLS] token, or via multiple representations, e.g., using an embedding for each token of the query and document (exemplified by ColBERT). In this work, we conduct the first study into the potential for multiple representation dense retrieval to be enhanced using pseudo-relevance feedback and present our proposed approach ColBERT-PRF. In particular, based on the pseudo-relevant set of documents identified using a first-pass dense retrieval, ColBERT-PRF extracts the representative feedback embeddings from the document embeddings of the pseudo-relevant set. Among the representative feedback embeddings, the embeddings that most highly discriminate among documents are employed as the expansion embeddings, which are then added to the original query representation. We show that these additional expansion embeddings both enhance the effectiveness of a reranking of the initial query results as well as an additional dense retrieval operation. Indeed, experiments on the MSMARCO passage ranking dataset show that MAP can be improved by up to 26% on the TREC 2019 query set and 10% on the TREC 2020 query set by the application of our proposed ColBERT-PRF method on a ColBERT dense retrieval approach.We further validate the effectiveness of our proposed pseudo-relevance feedback technique for a dense retrieval model on MSMARCO document ranking and TREC Robust04 document ranking tasks. For instance, ColBERT-PRF exhibits up to 21% and 14% improvement in MAP over the ColBERT E2E model on the MSMARCO document ranking TREC 2019 and TREC 2020 query sets, respectively. Additionally, we study the effectiveness of variants of the ColBERT-PRF model with different weighting methods. Finally, we show that ColBERT-PRF can be made more efficient, attaining up to 4.54× speedup over the default ColBERT-PRF model, and with little impact on effectiveness, through the application of approximate scoring and different clustering methods.
23 November 2018 in Foundations and Trends® in Information Retrieval
Search engines are exceptionally important tools for accessing information in today’s world. In satisfying the information needs of millions of users, the effectiveness (the quality of the search results) and the efficiency (the speed at which the results are returned to the users) of a search engine are two goals that form a natural trade-off, as techniques that improve the effectiveness of the search engine can also make it less efficient. Meanwhile, search engines continue to rapidly evolve, with larger indexes, more complex retrieval strategies and growing query volumes. Hence, there is a need for the development of efficient query processing infrastructures that make appropriate sacrifices in effectiveness in order to make gains in efficiency. This survey comprehensively reviews the foundations of search engines, from index layouts to basic term-at-a-time (TAAT) and document-at-a-time (DAAT) query processing strategies, while also providing the latest trends in the literature in efficient query processing, including the coherent and systematic reviews of techniques such as dynamic pruning and impact-sorted posting lists as well as their variants and optimisations. Our explanations of query processing strategies, for instance the WAND and BMW dynamic pruning algorithms, are presented with illustrative figures showing how the processing state changes as the algorithms progress. Moreover, acknowledging the recent trends in applying a cascading infrastructure within search systems, this survey describes techniques for efficiently integrating effective learned models, such as those obtained from learning-to-rank techniques. The survey also covers the selective application of query processing techniques, often achieved by predicting the response times of the search engine (known as query efficiency prediction), and making per-query tradeoffs between efficiency and effectiveness to ensure that the required retrieval speed targets can be met. Finally, the survey concludes with a summary of open directions in efficient search infrastructures, namely the use of signatures, real-time, energy-efficient and modern hardware and software architectures.