The principal goal of the TREC Neural Cross-Language Information Retrieval (NeuCLIR) track is to study the impact of neural approaches to cross-language information retrieval. The track has created four collections, large collections of Chinese, Persian, and Russian newswire and a smaller collection of Chinese scientific abstracts. The principal tasks are ranked retrieval of news in one of the three languages, using English topics. Results for a multilingual task, also with English topics but with documents from all three newswire collections, are also reported. New in this second year of the track is a pilot technical documents CLIR task for ranked retrieval of Chinese technical documents using English topics. A total of 220 runs across all tasks were submitted by six participating teams and, as baselines, by track coordinators. Task descriptions and results are presented.
Transformer-based Cross-Encoders achieve state-of-the-art effectiveness in text retrieval. However, Cross-Encoders based on large transformer models (such as BERT or T5) are computationally expensive and allow for scoring only a small number of documents within a reasonably small latency window. However, keeping search latencies low is important for user satisfaction and energy usage. In this paper, we show that weaker shallow transformer models (i.e., transformers with a limited number of layers) actually perform better than full-scale models when constrained to these practical low-latency settings since they can estimate the relevance of more documents in the same time budget. We further show that shallow transformers may benefit from the generalized Binary Cross-Entropy (gBCE) training scheme, which has recently demonstrated success for recommendation tasks. Our experiments with TREC Deep Learning passage ranking query sets demonstrate significant improvements in shallow and full-scale models in low-latency scenarios. For example, when the latency limit is 25ms per query, MonoBERT-Large (a cross-encoder based on a full-scale BERT model) is only able to achieve NDCG@10 of 0.431 on TREC DL 2019, while TinyBERT-gBCE (a cross-encoder based on TinyBERT trained with gBCE) reaches NDCG@10 of 0.652, a +51% gain over MonoBERT-Large. We also show that shallow Cross-Encoders are effective even when used without a GPU (e.g., with CPU inference, NDCG@10 decreases only by 3% compared to GPU inference with 50ms latency), which makes Cross-Encoders practical to run even without specialized hardware acceleration.
Modern Large Language Models (LLMs) are capable of following long and complex instructions that enable a diverse amount of user tasks. However, despite Information Retrieval (IR) models using LLMs as the backbone of their architectures, nearly all of them still only take queries as input, with no instructions. For the handful of recent models that do take instructions, it's unclear how they use them. We introduce our dataset FollowIR, which contains a rigorous instruction evaluation benchmark as well as a training set for helping IR models learn to better follow real-world instructions. FollowIR builds off the long history of the TREC conferences: as TREC provides human annotators with instructions (also known as narratives) to determine document relevance, so should IR models be able to understand and decide relevance based on these detailed instructions. Our evaluation benchmark starts with three deeply judged TREC collections and alters the annotator instructions, re-annotating relevant documents. Through this process, we can measure how well IR models follow instructions, through a new pairwise evaluation framework. Our results indicate that existing retrieval models fail to correctly use instructions, using them for basic keywords and struggling to understand long-form information. However, we show that it is possible for IR models to learn to follow complex instructions: our new FollowIR-7B model has significant improvements (over 13%) after fine-tuning on our training set.
Modern sequence-to-sequence relevance models like monoT5 can effectively capture complex textual interactions between queries and documents through cross-encoding. However, the use of natural language tokens in prompts, such as Query, Document, and Relevant for monoT5, opens an attack vector for malicious documents to manipulate their relevance score through prompt injection, e.g., by adding target words such as true. Since such possibilities have not yet been considered in retrieval evaluation, we analyze the impact of query-independent prompt injection via manually constructed templates and LLM-based rewriting of documents on several existing relevance models. Our experiments on the TREC Deep Learning track show that adversarial documents can easily manipulate different sequence-to-sequence relevance models, while BM25 (as a typical lexical model) is not affected. Remarkably, the attacks also affect encoder-only relevance models (which do not rely on natural language prompt tokens), albeit to a lesser extent.
Information retrieval models have witnessed a paradigm shift from unsupervised statistical approaches to feature-based supervised approaches to completely data-driven ones that make use of the pre-training of large language models. While the increasing complexity of the search models have been able to demonstrate improvements in effectiveness (measured in terms of relevance of top-retrieved results), a question worthy of a thorough inspection is - "how explainable are these models?", which is what this paper aims to evaluate. In particular, we propose a common evaluation platform to systematically evaluate the explainability of any ranking model (the explanation algorithm being identical for all the models that are to be evaluated). In our proposed framework, each model, in addition to returning a ranked list of documents, also requires to return a list of explanation units or rationales for each document. This meta-information from each document is then used to measure how locally consistent these rationales are as an intrinsic measure of interpretability - one that does not require manual relevance assessments. Additionally, as an extrinsic measure, we compute how relevant these rationales are by leveraging sub-document level relevance assessments. Our findings show a number of interesting observations, such as sentence-level rationales are more consistent, an increase in complexity mostly leads to less consistent explanations, and that interpretability measures offer a complementary dimension of evaluation of IR systems because consistency is not well-correlated with nDCG at top ranks.
Pseudo-relevance feedback (PRF) can enhance average retrieval effectiveness over a sufficiently large number of queries. However, PRF often introduces a drift into the original information need, thus hurting the retrieval effectiveness of several queries. While a selective application of PRF can potentially alleviate this issue, previous approaches have largely relied on unsupervised or feature-based learning to determine whether a query should be expanded. In contrast, we revisit the problem of selective PRF from a deep learning perspective, presenting a model that is entirely data-driven and trained in an end-to-end manner. The proposed model leverages a transformer-based bi-encoder architecture. Additionally, to further improve retrieval effectiveness with this selective PRF approach, we make use of the model's confidence estimates to combine the information from the original and expanded queries. In our experiments, we apply this selective feedback on a number of different combinations of ranking and feedback models, and show that our proposed approach consistently improves retrieval effectiveness for both sparse and dense ranking models, with the feedback models being either sparse, dense or generative.
One advantage of neural ranking models is that they are meant to generalise well in situations of synonymity i.e. where two words have similar or identical meanings. In this paper, we investigate and quantify how well various ranking models perform in a clear-cut case of synonymity: when words are simply expressed in different surface forms due to regional differences in spelling conventions (e.g., color vs colour). We first explore the prevalence of American and British English spelling conventions in datasets used for the pre-training, training and evaluation of neural retrieval methods, and find that American spelling conventions are far more prevalent. Despite these biases in the training data, we find that retrieval models often generalise well in this case of synonymity. We explore the effect of document spelling normalisation in retrieval and observe that all models are affected by normalising the document's spelling. While they all experience a drop in performance when normalised to a different spelling convention than that of the query, we observe varied behaviour when the document is normalised to share the query spelling convention: lexical models show improvements, dense retrievers remain unaffected, and re-rankers exhibit contradictory behaviour.
Performing automatic reformulations of a user's query is a popular paradigm used in information retrieval (IR) for improving effectiveness -- as exemplified by the pseudo-relevance feedback approaches, which expand the query in order to alleviate the vocabulary mismatch problem. Recent advancements in generative language models have demonstrated their ability in generating responses that are relevant to a given prompt. In light of this success, we seek to study the capacity of such models to perform query reformulation and how they compare with long-standing query reformulation methods that use pseudo-relevance feedback. In particular, we investigate two representative query reformulation frameworks, GenQR and GenPRF. GenQR directly reformulates the user's input query, while GenPRF provides additional context for the query by making use of pseudo-relevance feedback information. For each reformulation method, we leverage different techniques, including fine-tuning and direct prompting, to harness the knowledge of language models. The reformulated queries produced by the generative models are demonstrated to markedly benefit the effectiveness of a state-of-the-art retrieval pipeline on four TREC test collections (varying from TREC 2004 Robust to the TREC 2019 Deep Learning). Furthermore, our results indicate that our studied generative models can outperform various statistical query expansion approaches while remaining comparable to other existing complex neural query reformulation models, with the added benefit of being simpler to implement.
Retrieval approaches that score documents based on learned dense vectors (i.e., dense retrieval) rather than lexical signals (i.e., conventional retrieval) are increasingly popular. Their ability to identify related documents that do not necessarily contain the same terms as those appearing in the user's query (thereby improving recall) is one of their key advantages. However, to actually achieve these gains, dense retrieval approaches typically require an exhaustive search over the document collection, making them considerably more expensive at query-time than conventional lexical approaches. Several techniques aim to reduce this computational overhead by approximating the results of a full dense retriever. Although these approaches reasonably approximate the top results, they suffer in terms of recall -- one of the key advantages of dense retrieval. We introduce 'LADR' (Lexically-Accelerated Dense Retrieval), a simple-yet-effective approach that improves the efficiency of existing dense retrieval models without compromising on retrieval effectiveness. LADR uses lexical retrieval techniques to seed a dense retrieval exploration that uses a document proximity graph. We explore two variants of LADR: a proactive approach that expands the search space to the neighbors of all seed documents, and an adaptive approach that selectively searches the documents with the highest estimated relevance in an iterative fashion. Through extensive experiments across a variety of dense retrieval models, we find that LADR establishes a new dense retrieval effectiveness-efficiency Pareto frontier among approximate k nearest neighbor techniques. Further, we find that when tuned to take around 8ms per query in retrieval latency on our hardware, LADR consistently achieves both precision and recall that are on par with an exhaustive search on standard benchmarks.
Sparse and dense pseudo-relevance feedback (PRF) approaches perform poorly on challenging queries due to low precision in first-pass retrieval. However, recent advances in neural language models (NLMs) can re-rank relevant documents to top ranks, even when few are in the re-ranking pool. This paper first addresses the problem of poor pseudo-relevance feedback by simply applying re-ranking prior to query expansion and re-executing this query. We find that this change alone can improve the retrieval effectiveness of sparse and dense PRF approaches by 5-8%. Going further, we propose a new expansion model, Latent Entity Expansion (LEE), a fine-grained word and entity-based relevance modelling incorporating localized features. Finally, we include an "adaptive" component to the retrieval process, which iteratively refines the re-ranking pool during scoring using the expansion model, i.e. we "re-rank - expand - repeat". Using LEE, we achieve (to our knowledge) the best NDCG, MAP and R@1000 results on the TREC Robust 2004 and CODEC adhoc document datasets, demonstrating a significant advancement in expansion effectiveness.