Achieving precise semantic control over the latent spaces of Variational AutoEncoders (VAEs) holds significant value for downstream tasks in NLP as the underlying generative mechanisms could be better localised, explained and improved upon. Recent research, however, has struggled to achieve consistent results, primarily due to the inevitable loss of semantic information in the variational bottleneck and limited control over the decoding mechanism. To overcome these challenges, we investigate discrete latent spaces in Vector Quantized Variational AutoEncoders (VQVAEs) to improve semantic control and generation in Transformer-based VAEs. In particular, We propose T5VQVAE, a novel model that leverages the controllability of VQVAEs to guide the self-attention mechanism in T5 at the token-level, exploiting its full generalization capabilities. Experimental results indicate that T5VQVAE outperforms existing state-of-the-art VAE models, including Optimus, in terms of controllability and preservation of semantic information across different tasks such as auto-encoding of sentences and mathematical expressions, text transfer, and inference. Moreover, T5VQVAE exhibits improved inference capabilities, suggesting potential applications for downstream natural language and symbolic reasoning tasks.
The language ability of Large Language Models (LLMs) is often unbalanced towards English because of the imbalance in the distribution of the pre-training data. This disparity is demanded in further fine-tuning and affecting the cross-lingual abilities of LLMs. In this paper, we propose to empower Instructiontuned LLMs (It-LLMs) in languages other than English by building semantic alignment between them. Hence, we propose CrossAlpaca, an It-LLM with cross-lingual instruction-following and Translation-following demonstrations to improve semantic alignment between languages. We validate our approach on the multilingual Question Answering (QA) benchmarks XQUAD and MLQA and adapted versions of MMLU and BBH. Our models, tested over six different languages, outperform the It-LLMs tuned on monolingual data. The final results show that instruction tuning on non-English data is not enough and that semantic alignment can be further improved by Translation-following demonstrations.
The derivation of mathematical results in specialised fields, using Large Language Models (LLMs), is an emerging research direction that can help identify models' limitations, and potentially support mathematical discovery. In this paper, we leverage a symbolic engine to generate derivations of equations at scale, and investigate the capabilities of LLMs when deriving goal equations from premises. Specifically, we employ in-context learning for GPT and fine-tune a range of T5 models to compare the robustness and generalisation of pre-training strategies to specialised models. Empirical results show that fine-tuned FLAN-T5-large (MathT5) outperforms GPT models on all static and out-of-distribution test sets in conventional scores. However, an in-depth analysis reveals that the fine-tuned models are more sensitive to perturbations involving unseen symbols and (to a lesser extent) changes to equation structure. In addition, we analyse 1.7K equations, and over 200 derivations, to highlight common reasoning errors such as the inclusion of incorrect, irrelevant, and redundant equations. Finally, we explore the suitability of existing metrics for evaluating mathematical derivations and find evidence that, while they can capture general properties such as sensitivity to perturbations, they fail to highlight fine-grained reasoning errors and essential differences between models. Overall, this work demonstrates that training models on synthetic data may improve their math capabilities beyond much larger LLMs, but current metrics are not appropriately assessing the quality of generated mathematical text.
Explainable natural language inference aims to provide a mechanism to produce explanatory (abductive) inference chains which ground claims to their supporting premises. A recent corpus called EntailmentBank strives to advance this task by explaining the answer to a question using an entailment tree \cite{dalvi2021explaining}. They employ the T5 model to directly generate the tree, which can explain how the answer is inferred. However, it lacks the ability to explain and control the generation of intermediate steps, which is crucial for the multi-hop inference process. % One recent corpus, EntailmentBank, aims to push this task forward by explaining an answer to a question according to an entailment tree \cite{dalvi2021explaining}. They employ T5 to generate the tree directly, which can explain how the answer is inferred but cannot explain how the intermediate is generated, which is essential to the multi-hop inference process. In this work, we focus on proposing a controlled natural language inference architecture for multi-premise explanatory inference. To improve control and enable explanatory analysis over the generation, we define lexical inference types based on Abstract Meaning Representation (AMR) graph and modify the architecture of T5 to learn a latent sentence representation (T5 bottleneck) conditioned on said type information. We also deliver a dataset of approximately 5000 annotated explanatory inference steps, with well-grounded lexical-symbolic operations. Experimental results indicate that the inference typing induced at the T5 bottleneck can help T5 to generate a conclusion under explicit control.
Inferring over and extracting information from Large Language Models (LLMs) trained on a large corpus of scientific literature can potentially drive a new era in biomedical research, reducing the barriers for accessing existing medical evidence. This work examines the potential of LLMs for dialoguing with biomedical background knowledge, using the context of antibiotic discovery as an exemplar motivational scenario. The context of biomedical discovery from natural products entails understanding the relational evidence between an organism, an associated chemical and its associated antibiotic properties. We provide a systematic assessment on the ability of LLMs to encode and express these relations, verifying for fluency, prompt-alignment, semantic coherence, factual knowledge and specificity of generated responses. The systematic analysis is applied to nine state-of-the-art models (including ChatGPT and GPT-4) in two prompting-based tasks: chemical compound definition generation and chemical compound-fungus relation determination. Results show that while recent models have improved in fluency, factual accuracy is still low and models are biased towards over-represented entities. The ability of LLMs to serve as biomedical knowledge bases is questioned, and the need for additional systematic evaluation frameworks is highlighted. The best performing GPT-4 produced a factual definition for 70% of chemical compounds and 43.6% factual relations to fungi, whereas the best open source model BioGPT-large 30% of the compounds and 30% of the relations for the best-performing prompt. The results show that while LLMs are currently not fit for purpose to be used as biomedical factual knowledge bases, there is a promising emerging property in the direction of factuality as the models become domain specialised, scale-up in size and level of human feedback.
Whether Transformers can learn to apply symbolic rules and generalise to out-of-distribution examples is an open research question. In this paper, we devise a data generation method for producing intricate mathematical derivations, and systematically perturb them with respect to syntax, structure, and semantics. Our task-agnostic approach generates equations, annotations, and inter-equation dependencies, employing symbolic algebra for scalable data production and augmentation. We then instantiate a general experimental framework on next-equation prediction, assessing systematic mathematical reasoning and generalisation of Transformer encoders on a total of 200K examples. The experiments reveal that perturbations heavily affect performance and can reduce F1 scores of $97\%$ to below $17\%$, suggesting that inference is dominated by surface-level patterns unrelated to a deeper understanding of mathematical operators. These findings underscore the importance of rigorous, large-scale evaluation frameworks for revealing fundamental limitations of existing models.
Large Language Models (LLMs) have exploded a new heatwave of AI, for their ability to engage end-users in human-level conversations with detailed and articulate answers across many knowledge domains. In response to their fast adoption in many industrial applications, this survey concerns their safety and trustworthiness. First, we review known vulnerabilities of the LLMs, categorising them into inherent issues, intended attacks, and unintended bugs. Then, we consider if and how the Verification and Validation (V&V) techniques, which have been widely developed for traditional software and deep learning models such as convolutional neural networks, can be integrated and further extended throughout the lifecycle of the LLMs to provide rigorous analysis to the safety and trustworthiness of LLMs and their applications. Specifically, we consider four complementary techniques: falsification and evaluation, verification, runtime monitoring, and ethical use. Considering the fast development of LLMs, this survey does not intend to be complete (although it includes 300 references), especially when it comes to the applications of LLMs in various domains, but rather a collection of organised literature reviews and discussions to support the quick understanding of the safety and trustworthiness issues from the perspective of V&V.
Rigorous evaluation of the causal effects of semantic features on language model predictions can be hard to achieve for natural language reasoning problems. However, this is such a desirable form of analysis from both an interpretability and model evaluation perspective, that it is valuable to zone in on specific patterns of reasoning with enough structure and regularity to be able to identify and quantify systematic reasoning failures in widely-used models. In this vein, we pick a portion of the NLI task for which an explicit causal diagram can be systematically constructed: in particular, the case where across two sentences (the premise and hypothesis), two related words/terms occur in a shared context. In this work, we apply causal effect estimation strategies to measure the effect of context interventions (whose effect on the entailment label is mediated by the semantic monotonicity characteristic) and interventions on the inserted word-pair (whose effect on the entailment label is mediated by the relation between these words.). Following related work on causal analysis of NLP models in different settings, we adapt the methodology for the NLI task to construct comparative model profiles in terms of robustness to irrelevant changes and sensitivity to impactful changes.
Probing strategies have been shown to detect the presence of various linguistic features in large language models; in particular, semantic features intermediate to the "natural logic" fragment of the Natural Language Inference task (NLI). In the case of natural logic, the relation between the intermediate features and the entailment label is explicitly known: as such, this provides a ripe setting for interventional studies on the NLI models' representations, allowing for stronger causal conjectures and a deeper critical analysis of interventional probing methods. In this work, we carry out new and existing representation-level interventions to investigate the effect of these semantic features on NLI classification: we perform amnesic probing (which removes features as directed by learned linear probes) and introduce the mnestic probing variation (which forgets all dimensions except the probe-selected ones). Furthermore, we delve into the limitations of these methods and outline some pitfalls have been obscuring the effectivity of interventional probing studies.