Abstract:Local feature selection in machine learning provides instance-specific explanations by focusing on the most relevant features for each prediction, enhancing the interpretability of complex models. However, such methods tend to produce misleading explanations by encoding additional information in their selections. In this work, we attribute the problem of misleading selections by formalizing the concepts of label and feature leakage. We rigorously derive the necessary and sufficient conditions under which we can guarantee no leakage, and show existing methods do not meet these conditions. Furthermore, we propose the first local feature selection method that is proven to have no leakage called SUWR. Our experimental results indicate that SUWR is less prone to overfitting and combines state-of-the-art predictive performance with high feature-selection sparsity. Our generic and easily extendable formal approach provides a strong theoretical basis for future work on interpretability with reliable explanations.
Abstract:Open-domain complex Question Answering (QA) is a difficult task with challenges in evidence retrieval and reasoning. The complexity of such questions could stem from questions being compositional, hybrid evidence, or ambiguity in questions. While retrieval performance for classical QA tasks is well explored, their capabilities for heterogeneous complex retrieval tasks, especially in an open-domain setting, and the impact on downstream QA performance, are relatively unexplored. To address this, in this work, we propose a benchmark composing diverse complex QA tasks and provide a toolkit to evaluate state-of-the-art pre-trained dense and sparse retrieval models in an open-domain setting. We observe that late interaction models and surprisingly lexical models like BM25 perform well compared to other pre-trained dense retrieval models. In addition, since context-based reasoning is critical for solving complex QA tasks, we also evaluate the reasoning capabilities of LLMs and the impact of retrieval performance on their reasoning capabilities. Through experiments, we observe that much progress is to be made in retrieval for complex QA to improve downstream QA performance. Our software and related data can be accessed at https://github.com/VenkteshV/DEXTER
Abstract:Experimental control involves a lot of manual effort with non-trivial decisions for precise adjustments. Here, we study the automatic experimental alignment for coupling laser light into an optical fiber using reinforcement learning (RL). We face several real-world challenges, such as time-consuming training, partial observability, and noisy actions due to imprecision in the mirror steering motors. We show that we can overcome these challenges: To save time, we use a virtual testbed to tune our environment for dealing with partial observability and use relatively sample-efficient model-free RL algorithms like Soft Actor-Critic (SAC) or Truncated Quantile Critics (TQC). Furthermore, by fully training on the experiment, the agent learns directly to handle the noise present. In our extensive experimentation, we show that we are able to achieve 90% coupling, showcasing the effectiveness of our proposed approaches. We reach this efficiency, which is comparable to that of a human expert, without additional feedback loops despite the motors' inaccuracies. Our result is an example of the readiness of RL for real-world tasks. We consider RL a promising tool for reducing the workload in labs.
Abstract:Neural ranking models have become increasingly popular for real-world search and recommendation systems in recent years. Unlike their tree-based counterparts, neural models are much less interpretable. That is, it is very difficult to understand their inner workings and answer questions like how do they make their ranking decisions? or what document features do they find important? This is particularly disadvantageous since interpretability is highly important for real-world systems. In this work, we explore feature selection for neural learning-to-rank (LTR). In particular, we investigate six widely-used methods from the field of interpretable machine learning (ML) and introduce our own modification, to select the input features that are most important to the ranking behavior. To understand whether these methods are useful for practitioners, we further study whether they contribute to efficiency enhancement. Our experimental results reveal a large feature redundancy in several LTR benchmarks: the local selection method TabNet can achieve optimal ranking performance with less than 10 features; the global methods, particularly our G-L2X, require slightly more selected features, but exhibit higher potential in improving efficiency. We hope that our analysis of these feature selection methods will bring the fields of interpretable ML and LTR closer together.
Abstract:An important problem in text-ranking systems is handling the hard queries that form the tail end of the query distribution. The difficulty may arise due to the presence of uncommon, underspecified, or incomplete queries. In this work, we improve the ranking performance of hard or difficult queries without compromising the performance of other queries. Firstly, we do LLM based query enrichment for training queries using relevant documents. Next, a specialized ranker is fine-tuned only on the enriched hard queries instead of the original queries. We combine the relevance scores from the specialized ranker and the base ranker, along with a query performance score estimated for each query. Our approach departs from existing methods that usually employ a single ranker for all queries, which is biased towards easy queries, which form the majority of the query distribution. In our extensive experiments on the DL-Hard dataset, we find that a principled query performance based scoring method using base and specialized ranker offers a significant improvement of up to 25% on the passage ranking task and up to 48.4% on the document ranking task when compared to the baseline performance of using original queries, even outperforming SOTA model.
Abstract:Automated fact checking has gained immense interest to tackle the growing misinformation in the digital era. Existing systems primarily focus on synthetic claims on Wikipedia, and noteworthy progress has also been made on real-world claims. In this work, we release Numtemp, a diverse, multi-domain dataset focused exclusively on numerical claims, encompassing temporal, statistical and diverse aspects with fine-grained metadata and an evidence collection without leakage. This addresses the challenge of verifying real-world numerical claims, which are complex and often lack precise information, not addressed by existing works that mainly focus on synthetic claims. We evaluate and quantify the limitations of existing solutions for the task of verifying numerical claims. We also evaluate claim decomposition based methods, numerical understanding based models and our best baselines achieves a macro-F1 of 58.32. This demonstrates that Numtemp serves as a challenging evaluation set for numerical claim verification.
Abstract:Feature attributions are a commonly used explanation type, when we want to posthoc explain the prediction of a trained model. Yet, they are not very well explored in IR. Importantly, feature attribution has rarely been rigorously defined, beyond attributing the most important feature the highest value. What it means for a feature to be more important than others is often left vague. Consequently, most approaches focus on just selecting the most important features and under utilize or even ignore the relative importance within features. In this work, we rigorously define the notion of feature attribution for ranking models, and list essential properties that a valid attribution should have. We then propose RankingSHAP as a concrete instantiation of a list-wise ranking attribution method. Contrary to current explanation evaluation schemes that focus on selections, we propose two novel evaluation paradigms for evaluating attributions over learning-to-rank models. We evaluate RankingSHAP for commonly used learning-to-rank datasets to showcase the more nuanced use of an attribution method while highlighting the limitations of selection-based explanations. In a simulated experiment we design an interpretable model to demonstrate how list-wise ranking attributes can be used to investigate model decisions and evaluate the explanations qualitatively. Because of the contrastive nature of the ranking task, our understanding of ranking model decisions can substantially benefit from feature attribution explanations like RankingSHAP.
Abstract:Large language models (LLMs) have recently gained significant attention due to their unparalleled ability to perform various natural language processing tasks. These models, benefiting from their advanced natural language understanding capabilities, have demonstrated impressive zero-shot performance. However, the pre-training data utilized in LLMs is often confined to a specific corpus, resulting in inherent freshness and temporal scope limitations. Consequently, this raises concerns regarding the effectiveness of LLMs for tasks involving temporal intents. In this study, we aim to investigate the underlying limitations of general-purpose LLMs when deployed for tasks that require a temporal understanding. We pay particular attention to handling factual temporal knowledge through three popular temporal QA datasets. Specifically, we observe low performance on detailed questions about the past and, surprisingly, for rather new information. In manual and automatic testing, we find multiple temporal errors and characterize the conditions under which QA performance deteriorates. Our analysis contributes to understanding LLM limitations and offers valuable insights into developing future models that can better cater to the demands of temporally-oriented tasks. The code is available\footnote{https://github.com/jwallat/temporalblindspots}.
Abstract:Contextual ranking models have delivered impressive performance improvements over classical models in the document ranking task. However, these highly over-parameterized models tend to be data-hungry and require large amounts of data even for fine-tuning. In this paper, we propose data-augmentation methods for effective and robust ranking performance. One of the key benefits of using data augmentation is in achieving sample efficiency or learning effectively when we have only a small amount of training data. We propose supervised and unsupervised data augmentation schemes by creating training data using parts of the relevant documents in the query-document pairs. We then adapt a family of contrastive losses for the document ranking task that can exploit the augmented data to learn an effective ranking model. Our extensive experiments on subsets of the MS MARCO and TREC-DL test sets show that data augmentation, along with the ranking-adapted contrastive losses, results in performance improvements under most dataset sizes. Apart from sample efficiency, we conclusively show that data augmentation results in robust models when transferred to out-of-domain benchmarks. Our performance improvements in in-domain and more prominently in out-of-domain benchmarks show that augmentation regularizes the ranking model and improves its robustness and generalization capability.
Abstract:Dual-encoder-based dense retrieval models have become the standard in IR. They employ large Transformer-based language models, which are notoriously inefficient in terms of resources and latency. We propose Fast-Forward indexes -- vector forward indexes which exploit the semantic matching capabilities of dual-encoder models for efficient and effective re-ranking. Our framework enables re-ranking at very high retrieval depths and combines the merits of both lexical and semantic matching via score interpolation. Furthermore, in order to mitigate the limitations of dual-encoders, we tackle two main challenges: Firstly, we improve computational efficiency by either pre-computing representations, avoiding unnecessary computations altogether, or reducing the complexity of encoders. This allows us to considerably improve ranking efficiency and latency. Secondly, we optimize the memory footprint and maintenance cost of indexes; we propose two complementary techniques to reduce the index size and show that, by dynamically dropping irrelevant document tokens, the index maintenance efficiency can be improved substantially. We perform evaluation to show the effectiveness and efficiency of Fast-Forward indexes -- our method has low latency and achieves competitive results without the need for hardware acceleration, such as GPUs.