Large language models (LLMs) have demonstrated strong reasoning abilities when prompted to generate chain-of-thought (CoT) explanations alongside answers. However, previous research on evaluating LLMs has solely focused on answer accuracy, neglecting the correctness of the generated CoT. In this paper, we delve deeper into the CoT reasoning capabilities of LLMs in multi-hop question answering by utilizing knowledge graphs (KGs). We propose a novel discriminative and generative CoT evaluation paradigm to assess LLMs’ knowledge of reasoning and the accuracy of the generated CoT. Through experiments conducted on 5 different families of LLMs across 2 multi-hop question-answering datasets, we find that LLMs possess sufficient knowledge to perform reasoning. However, there exists a significant disparity between answer accuracy and faithfulness of the CoT generated by LLMs, indicating that they often arrive at correct answers through incorrect reasoning.

Recent work has shown that an answer verification step introduced in Transformer-based answer selection models can significantly improve the state of the art in Question Answering. This step is performed by aggregating the embeddings of top k answer candidates to support the verification of a target answer. Although the approach is intuitive and sound, it still shows two limitations: (i) the supporting candidates are ranked only according to the relevancy with the question and not with the answer, and (ii) the support provided by the other answer candidates is suboptimal as these are retrieved independently of the target answer. In this paper, we address both drawbacks by proposing (i) a double reranking model, which, for each target answer, selects the best support; and (ii) a second neural retrieval stage designed to encode question and answer pair as the query, which finds more specific verification information. The results on well-known datasets for Answer Sentence Selection show significant improvement over the state of the art.

In very recent years more attention has been placed on probing the role of pre-training data in Large Language Models (LLMs) downstream behaviour. Despite the importance, there is no public tool that supports such analysis of pre-training corpora at large scale. To help research in this space, we launch Koala, a searchable index over large pre-training corpora using lossless compressed suffix arrays with highly efficient compression rate and search support. In its first release we index the public proportion of OPT 175B, GPT-3, GPT-Neo, GPT-Neo, LLaMA, BERT, ELECTRA, RoBERTA, XLNet pre-training corpora. Koala provides a framework to do forensic analysis on the current and future benchmarks as well as to assess the degree of memorization in the output from the LLMs. Koala is available for public use at https://koala-index.erc.monash.edu/.

This research studies graph-based approaches for Answer Sentence Selection (AS2), an essential component for retrieval-based Question Answering (QA) systems. During offline learning, our model constructs a small-scale relevant training graph per question in an unsupervised manner, and integrates with Graph Neural Networks. Graph nodes are question sentence to answer sentence pairs. We train and integrate state-of-the-art (SOTA) models for computing scores between question-question, question-answer, and answer-answer pairs, and use thresholding on relevance scores for creating graph edges. Online inference is then performed to solve the AS2 task on unseen queries. Experiments on two well-known academic benchmarks and a real-world dataset show that our approach consistently outperforms SOTA QA baseline models.

Generalising to unseen domains is under-explored and remains a challenge in neural machine translation. Inspired by recent research in parameter-efficient transfer learning from pretrained models, this paper proposes a fusion-based generalisation method that learns to combine domain-specific parameters. We propose a leave-one-domain-out training strategy to avoid information leaking to address the challenge of not knowing the test domain during training time. Empirical results on three language pairs show that our proposed fusion method outperforms other baselines up to +0.8 BLEU score on average.

We introduce question answering with a cotext in focus, a task that simulates a free interaction with a QA system. The user reads on a screen some information about a topic, and they can follow-up with questions that can be either related or not to the topic; and the answer can be found in the document containing the screen content or from other pages. We call such information context. To study the task, we construct FocusQA, a dataset for answer sentence selection (AS2) with 12,165011unique question/context pairs, and a total of 109,940 answers. To build the dataset, we developed a novel methodology that takes existing questions and pairs them with relevant contexts. To show the benefits of this approach, we present a comparative analysis with a set of questions written by humans after reading the context, showing that our approach greatly helps in eliciting more realistic question/context pairs. Finally, we show that the task poses several challenges for incorporating contextual information. In this respect, we introduce strong baselines for answer sentence selection that outperform the precision of state-of-the-art models for AS2 up to 21.3% absolute points.

This paper studies joint models for selecting correct answer sentences among the top k provided by answer sentence selection (AS2) modules, which are core components of retrieval-based Question Answering (QA) systems. Our work shows that a critical step to effectively exploiting an answer set regards modeling the interrelated information between pair of answers. For this purpose, we build a three-way multi-classifier, which decides if an answer supports, refutes, or is neutral with respect to another one. More specifically, our neural architecture integrates a state-of-the-art AS2 module with the multi-classifier, and a joint layer connecting all components. We tested our models on WikiQA, TREC-QA, and a real-world dataset. The results show that our models obtain the new state of the art in AS2.

We introduce AVA, an automatic evaluation approach for Question Answering, which given a set of questions associated with Gold Standard answers (references), can estimate system Accuracy. AVA uses Transformer-based language models to encode question, answer, and reference texts. This allows for effectively assessing answer correctness using similarity between the reference and an automatic answer, biased towards the question semantics. To design, train, and test AVA, we built multiple large training, development, and test sets on public and industrial benchmarks. Our innovative solutions achieve up to 74.7% F1 score in predicting human judgment for single answers. Additionally, AVA can be used to evaluate the overall system Accuracy with an error lower than 7% at 95% of confidence when measured on several QA systems.

We introduce a Content-based Document Alignment approach (CDA), an efficient method to align multilingual web documents based on content in creating parallel training data for machine translation (MT) systems operating at the industrial level. CDA works in two steps: (i) projecting documents of a web domain to a shared multilingual space; then (ii) aligning them based on the similarity of their representations in such space. We leverage lexical translation models to build vector representations using TF×IDF. CDA achieves performance comparable with state-of-the-art systems in the WMT-16 Bilingual Document Alignment Shared Task benchmark while operating in multilingual space. Besides, we created two web-scale datasets to examine the robustness of CDA in an industrial setting involving up to 28 languages and millions of documents. The experiments show that CDA is robust, cost-effective, and is significantly superior in (i) processing large and noisy web data and (ii) scaling to new and low-resourced languages.

Answer Sentence Selection (AS2) models are core components of efficient retrieval-based Question Answering (QA) systems. We present the Reference-based Weak Supervision (RWS), a fully automatic large-scale data pipeline that harvests high-quality weakly- supervised answer sentences from Web data, only requiring a question-reference pair as input. We evaluated the quality of the RWS-derived data by training TANDA models, which are the state of the art for AS2. Our results show that the data consistently bolsters TANDA on three different datasets. In particular, we set the new state of the art for AS2 to P@1=90.1%, and MAP=92.9%, on WikiQA. We record similar performance gains of RWS on a much larger dataset named Web-based Question Answering (WQA).

This paper considers the unsupervised domain adaptation problem for neural machine translation (NMT), where we assume the access to only monolingual text in either the source or target language in the new domain. We propose a cross-lingual data selection method to extract in-domain sentences in the missing language side from a large generic monolingual corpus. Our proposed method trains an adaptive layer on top of multilingual BERT by contrastive learning to align the representation between the source and target language. This then enables the transferability of the domain classifier between the languages in a zero-shot manner. Once the in-domain data is detected by the classifier, the NMT model is then adapted to the new domain by jointly learning translation and domain discrimination tasks. We evaluate our cross-lingual data selection method on NMT across five diverse domains in three language pairs, as well as a real-world scenario of translation for COVID-19. The results show that our proposed method outperforms other selection baselines up to +1.5 BLEU score.

Recent work has shown the importance of adaptation of broad-coverage contextualised embedding models on the domain of the target task of interest. Current self-supervised adaptation methods are simplistic, as the training signal comes from a small percentage of randomly masked-out tokens. In this paper, we show that careful masking strategies can bridge the knowledge gap of masked language models (MLMs) about the domains more effectively by allocating self-supervision where it is needed. Furthermore, we propose an effective training strategy by adversarially masking out those tokens which are harder to reconstruct by the underlying MLM. The adversarial objective leads to a challenging combinatorial optimisation problem over subsets of tokens, which we tackle efficiently through relaxation to a variational lowerbound and dynamic programming. On six unsupervised domain adaptation tasks involving named entity recognition, our method strongly outperforms the random masking strategy and achieves up to +1.64 F1 score improvements.

Heuristic-based active learning (AL) methods are limited when the data distribution of the underlying learning problems vary. Recent data-driven AL policy learning methods are also restricted to learn from closely related domains. We introduce a new sample-efficient method that learns the AL policy directly on the target domain of interest by using wake and dream cycles. Our approach interleaves between querying the annotation of the selected datapoints to update the underlying student learner and improving AL policy using simulation where the current student learner acts as an imperfect annotator. We evaluate our method on cross-domain and cross-lingual text classification and named entity recognition tasks. Experimental results show that our dream-based AL policy training strategy is more effective than applying the pretrained policy without further fine-tuning and better than the existing strong baseline methods that use heuristics or reinforcement learning.

Automated Post-Editing (PE) is the task of automatically correct common and repetitive errors found in machine translation (MT) output. In this paper, we present a neural programmer-interpreter approach to this task, resembling the way that human perform post-editing using discrete edit operations, wich we refer to as programs. Our model outperforms previous neural models for inducing PE programs on the WMT17 APE task for German-English up to +1 BLEU score and -0.7 TER scores.