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.

Previous studies have relied on existing question-answering benchmarks to evaluate the knowledge stored in large language models (LLMs). However, this approach has limitations regarding factual knowledge coverage, as it mostly focuses on generic domains which may overlap with the pretraining data. This paper proposes a framework to systematically assess the factual knowledge of LLMs by leveraging knowledge graphs (KGs). Our framework automatically generates a set of questions and expected answers from the facts stored in a given KG, and then evaluates the accuracy of LLMs in answering these questions. We systematically evaluate the state-of-the-art LLMs with KGs in generic and specific domains. The experiment shows that ChatGPT is consistently the top performer across all domains. We also find that LLMs performance depends on the instruction finetuning, domain and question complexity and is prone to adversarial context.

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.

Previous works mostly focus on either multilingual or multi-domain aspects of neural machine translation (NMT). This paper investigates whether the domain information can be transferred across languages on the composition of multi-domain and multilingual NMT, particularly for the incomplete data condition where in-domain bitext is missing for some language pairs. Our results in the curated leave-one-domain-out experiments show that multi-domain multilingual (MDML) NMT can boost zero-shot translation performance up to +10 gains on BLEU, as well as aid the generalisation of multi-domain NMT to the missing domain. We also explore strategies for effective integration of multilingual and multi-domain NMT, including language and domain tag combination and auxiliary task training. We find that learning domain-aware representations and adding target-language tags to the encoder leads to effective MDML-NMT.

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.

Interpretability and explainability of deep neural net models are always challenging due to their size and complexity. Many previous works focused on visualizing internal components of neural networks to represent them through human-friendly concepts. On the other hand, in real life, when making a decision, human tends to rely on similar situations in the past. Thus, we argue that one potential approach to make the model interpretable and explainable is to design it in a way such that the model explicitly connects the current sample with the seen samples, and bases its decision on these samples. In this work, we design one such model: an explainable, evidence-based memory network architecture, which learns to summarize the dataset and extract supporting evidences to make its decision. The model achieves state-of-the-art performance on two popular question answering datasets, the TrecQA dataset and the WikiQA dataset. Via further analysis, we showed that this model can reliably trace the errors it has made in the validation step to the training instances that might have caused this error. We believe that this error-tracing capability might be beneficial in improving dataset quality in many applications.

Knowledge graph embedding methods often suffer from a limitation of memorizing valid triples to predict new ones for triple classification and search personalization problems. To this end, we introduce a novel embedding model, named R-MeN, that explores a relational memory network to encode potential dependencies in relationship triples. R-MeN considers each triple as a sequence of 3 input vectors that recurrently interact with a memory using a transformer self-attention mechanism. Thus R-MeN encodes new information from interactions between the memory and each input vector to return a corresponding vector. Consequently, R-MeN feeds these 3 returned vectors to a convolutional neural network-based decoder to produce a scalar score for the triple. Experimental results show that our proposed R-MeN obtains state-of-the-art results on SEARCH17 for the search personalization task, and on WN11 and FB13 for the triple classification task.

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.

In this paper, we introduce an embedding model, named CapsE, exploring a capsule network to model relationship triples (subject, relation, object). Our CapsE represents each triple as a 3-column matrix where each column vector represents the embedding of an element in the triple. This 3-column matrix is then fed to a convolution layer where multiple filters are operated to generate different feature maps. These feature maps are reconstructed into corresponding capsules which are then routed to another capsule to produce a continuous vector. The length of this vector is used to measure the plausibility score of the triple. Our proposed CapsE obtains better performance than previous state-of-the-art embedding models for knowledge graph completion on two benchmark datasets WN18RR and FB15k-237, and outperforms strong search personalization baselines on SEARCH17.

In this paper, we propose a novel embedding model, named ConvKB, for knowledge base completion. Our model ConvKB advances state-of-the-art models by employing a convolutional neural network, so that it can capture global relationships and transitional characteristics between entities and relations in knowledge bases. In ConvKB, each triple (head entity, relation, tail entity) is represented as a 3-column matrix where each column vector represents a triple element. This 3-column matrix is then fed to a convolution layer where multiple filters are operated on the matrix to generate different feature maps. These feature maps are then concatenated into a single feature vector representing the input triple. The feature vector is multiplied with a weight vector via a dot product to return a score. This score is then used to predict whether the triple is valid or not. Experiments show that ConvKB achieves better link prediction performance than previous state-of-the-art embedding models on two benchmark datasets WN18RR and FB15k-237.