Fact verification on tabular evidence incentivises the use of symbolic reasoning models where a logical form is constructed (e.g. a LISP-style program), providing greater verifiability than fully neural approaches. However, these systems typically rely on well-formed tables, restricting their use in many scenarios. An emerging symbolic reasoning paradigm for textual evidence focuses on natural logic inference, which constructs proofs by modelling set-theoretic relations between a claim and its evidence in natural language. This approach provides flexibility and transparency but is less compatible with tabular evidence since the relations do not extend to arithmetic functions. We propose a set-theoretic interpretation of numerals and arithmetic functions in the context of natural logic, enabling the integration of arithmetic expressions in deterministic proofs. We leverage large language models to generate arithmetic expressions by generating questions about salient parts of a claim which are answered by executing appropriate functions on tables. In a few-shot setting on FEVEROUS, we achieve an accuracy of 71.4, outperforming both fully neural and symbolic reasoning models by 3.4 points. When evaluated on TabFact without any further training, our method remains competitive with an accuracy lead of 0.5 points.

The recent development of fact verification systems with natural logic has enhanced their explainability by aligning claims with evidence through set-theoretic operators, providing faithful justifications. Despite these advancements, such systems often rely on a large amount of training data annotated with natural logic. To address this issue, we propose a zero-shot method that utilizes the generalization capabilities of instruction-tuned large language models. To comprehensively assess the zero-shot capabilities of our method and other fact verification systems, we evaluate all models on both artificial and real-world claims, including multilingual datasets. We also compare our method against other fact verification systems in two setups. First, in the zero-shot generalization setup, we demonstrate that our approach outperforms other systems that were not specifically trained on natural logic data, achieving an average accuracy improvement of 8.96 points over the best-performing baseline. Second, in the zero-shot transfer setup, we show that current systems trained on natural logic data do not generalize well to other domains, and our method outperforms these systems across all datasets with real-world claims.

Large Language Models have revolutionized various fields and industries, such as Conversational AI, Content Generation, Information Retrieval, Business Intelligence, and Medical, to name a few. One major application in the field of medical is to analyze and investigate clinical trials for entailment tasks.However, It has been observed that Large Language Models are susceptible to shortcut learning, factual inconsistency, and performance degradation with little variation in context. Adversarial and robust testing is performed to ensure the integrity of models output. But, ambiguity still persists. In order to ensure the integrity of the reasoning performed and investigate the model has correct syntactic and semantic understanding probing is used. Here, I used mnestic probing to investigate the Sci-five model, trained on clinical trial. I investigated the model for feature learnt with respect to natural logic. To achieve the target, I trained task specific probes. Used these probes to investigate the final layers of trained model. Then, fine tuned the trained model using iterative null projection. The results shows that model accuracy improved. During experimentation, I observed that size of the probe has affect on the fine tuning process.

In the interest of interpreting neural NLI models and their reasoning strategies, we carry out a systematic probing study which investigates whether these models capture the crucial semantic features central to natural logic: monotonicity and concept inclusion. Correctly identifying valid inferences in downward-monotone contexts is a known stumbling block for NLI performance, subsuming linguistic phenomena such as negation scope and generalized quantifiers. To understand this difficulty, we emphasize monotonicity as a property of a context and examine the extent to which models capture monotonicity information in the contextual embeddings which are intermediate to their decision making process. Drawing on the recent advancement of the probing paradigm, we compare the presence of monotonicity features across various models. We find that monotonicity information is notably weak in the representations of popular NLI models which achieve high scores on benchmarks, and observe that previous improvements to these models based on fine-tuning strategies have introduced stronger monotonicity features together with their improved performance on challenge sets.

Reasoning has been a central topic in artificial intelligence from the beginning. The recent progress made on distributed representation and neural networks continues to improve the state-of-the-art performance of natural language inference. However, it remains an open question whether the models perform real reasoning to reach their conclusions or rely on spurious correlations. Adversarial attacks have proven to be an important tool to help evaluate the Achilles' heel of the victim models. In this study, we explore the fundamental problem of developing attack models based on logic formalism. We propose NatLogAttack to perform systematic attacks centring around natural logic, a classical logic formalism that is traceable back to Aristotle's syllogism and has been closely developed for natural language inference. The proposed framework renders both label-preserving and label-flipping attacks. We show that compared to the existing attack models, NatLogAttack generates better adversarial examples with fewer visits to the victim models. The victim models are found to be more vulnerable under the label-flipping setting. NatLogAttack provides a tool to probe the existing and future NLI models' capacity from a key viewpoint and we hope more logic-based attacks will be further explored for understanding the desired property of reasoning.

We explore end-to-end trained differentiable models that integrate natural logic with neural networks, aiming to keep the backbone of natural language reasoning based on the natural logic formalism while introducing subsymbolic vector representations and neural components. The proposed model adapts module networks to model natural logic operations, which is enhanced with a memory component to model contextual information. Experiments show that the proposed framework can effectively model monotonicity-based reasoning, compared to the baseline neural network models without built-in inductive bias for monotonicity-based reasoning. Our proposed model shows to be robust when transferred from upward to downward inference. We perform further analyses on the performance of the proposed model on aggregation, showing the effectiveness of the proposed subcomponents on helping achieve better intermediate aggregation performance.

The increasing availability of large text corpora holds the promise of acquiring an unprecedented amount of knowledge from this text. However, current techniques are either specialized to particular domains or do not scale to large corpora. This dissertation develops a new technique for learning open-domain knowledge from unstructured web-scale text corpora. A first application aims to capture common sense facts: given a candidate statement about the world and a large corpus of known facts, is the statement likely to be true? We appeal to a probabilistic relaxation of natural logic -- a logic which uses the syntax of natural language as its logical formalism -- to define a search problem from the query statement to its appropriate support in the knowledge base over valid (or approximately valid) logical inference steps.

Recent work on entailment suggests that natural logics are well-suited to determining whether one sentence lexically entails another. We show how the EPILOG reasoning engine, designed for a natural language-like meaning representation (Episodic Logic, or EL), can be used to emulate natural logic inferences, while also enabling more general inferences such as ones from multiple premises, or ones based on world knowledge. Thus, to exploit the capabilities of EPILOG, we are working to populate its knowledge base with the kinds of lexical knowledge on which natural logics rely.