Chunhua Liu

Hong Yi Lin, Patanamon Thongtanunam, Christoph Treude et al.

Modern code review is a ubiquitous software quality assurance process aimed at identifying potential issues within newly written code. Despite its effectiveness, the process demands large amounts of effort from the human reviewers involved. To help alleviate this workload, researchers have trained deep learning models to imitate human reviewers in providing natural language code reviews. Formally, this task is known as code review comment generation. Prior work has demonstrated improvements in this task by leveraging machine learning techniques and neural models, such as transfer learning and the transformer architecture. However, the quality of the model generated reviews remain sub-optimal due to the quality of the open-source code review data used in model training. This is in part due to the data obtained from open-source projects where code reviews are conducted in a public forum, and reviewers possess varying levels of software development experience, potentially affecting the quality of their feedback. To accommodate for this variation, we propose a suite of experience-aware training methods that utilise the reviewers' past authoring and reviewing experiences as signals for review quality. Specifically, we propose experience-aware loss functions (ELF), which use the reviewers' authoring and reviewing ownership of a project as weights in the model's loss function. Through this method, experienced reviewers' code reviews yield larger influence over the model's behaviour. Compared to the SOTA model, ELF was able to generate higher quality reviews in terms of accuracy, informativeness, and comment types generated. The key contribution of this work is the demonstration of how traditional software engineering concepts such as reviewer experience can be integrated into the design of AI-based automated code review models.

Hong Yi Lin, Chunhua Liu, Haoyu Gao et al.

In today's AI-assisted software engineering landscape, developers increasingly depend on LLMs that are highly capable, yet inherently imperfect. The tendency of these models to produce incorrect outputs can reduce developer productivity. To this end, a canonical mitigation method is to provide calibrated confidence scores that faithfully reflect their likelihood of correctness at the instance-level. Such information allows users to make immediate decisions regarding output acceptance, abstain error-prone outputs, and better align their expectations with the model's capabilities. Since post-trained LLMs do not inherently produce well-calibrated confidence scores, researchers have developed post-hoc calibration methods, with global Platt-scaling of sequence-level confidence scores proving effective in many generative software engineering tasks but remaining unreliable or unexplored for automated code revision (ACR) tasks such as program repair, vulnerability repair, and code refinement. We hypothesise that the coarse-grained nature of this conventional method makes it ill-suited for ACR tasks, where correctness is often determined by local edit decisions and miscalibration can be sample-dependent, thereby motivating fine-grained confidence calibration. To address this, our study proposes local Platt-scaling applied separately to three different fine-grained confidence scores. Through experiments across 3 separate tasks and correctness metrics, as well as 14 different models of various sizes, we find that fine-grained confidence scores consistently achieve lower calibration error across a broader range of probability intervals, and this effect is further amplified when global Platt-scaling is applied. Our proposed approaches offer a practical solution to eliciting well-calibrated confidence scores, enabling more trustworthy and streamlined usage of imperfect models in ACR tasks.

Hong Yi Lin, Chunhua Liu, Haoyu Gao et al.

State-of-the-art large language models (LLMs) have demonstrated impressive code generation capabilities but struggle with real-world software engineering tasks, such as revising source code to address code reviews, hindering their practical use. Code review comments are often implicit, ambiguous, and colloquial, requiring models to grasp both code and human intent. This challenge calls for evaluating large language models' ability to bridge both technical and conversational contexts. While existing work has employed the automated code refinement (ACR) task to resolve these comments, current evaluation methods fall short, relying on text matching metrics that provide limited insight into model failures and remain susceptible to training data contamination. To address these limitations, we introduce a novel evaluation benchmark, $\textbf{CodeReviewQA}$ that enables us to conduct fine-grained assessment of model capabilities and mitigate data contamination risks. In CodeReviewQA, we decompose the generation task of code refinement into $\textbf{three essential reasoning steps}$: $\textit{change type recognition}$ (CTR), $\textit{change localisation}$ (CL), and $\textit{solution identification}$ (SI). Each step is reformulated as multiple-choice questions with varied difficulty levels, enabling precise assessment of model capabilities, while mitigating data contamination risks. Our comprehensive evaluation spans 72 recently released large language models on $\textbf{900 manually curated, high-quality examples}$ across nine programming languages. Our results show that CodeReviewQA is able to expose specific model weaknesses in code review comprehension, disentangled from their generative automated code refinement results.

Chunhua Liu, Kabir Manandhar Shrestha, Sukai Huang

As large language models (LLMs) increasingly mediate cross-cultural communication, their behavior still reflects the distributional bias of the languages and viewpoints that are over-represented in their pre-training corpora. Yet, it remains a challenge to model and align culture due to limited cultural knowledge and a lack of exploration into effective learning approaches. We introduce a cost-efficient, cognitively grounded remedy: parameter-efficient fine-tuning on native speakers' free word-association norms, which encode implicit cultural schemas. Leveraging English-US and Mandarin associations from the Small-World-of-Words project, we adapt Llama-3.1-8B and Qwen-2.5-7B via supervised fine-tuning (SFT) and PPO-based preference optimization. SFT boosts held-out association Precision at 5 by 16-20% in English and 43-165% in Mandarin, lifts median concreteness by +0.20, and attains human-level valence and arousal. These lexical gains transfer: on World-Values-Survey questions, fine-tuned models shift answer distributions toward the target culture, and on a 50-item high-tension subset, Qwen's Chinese-aligned responses double while Llama's US bias drops by one-third. Our 7-8B models rival or beat vanilla 70B baselines, showing that a few million culture-grounded associations can instill value alignment without costly retraining. Our work highlights both the promise and the need for future research grounded in human cognition in improving cultural alignment in AI models.

Linh Nguyen, Chunhua Liu, Hong Yi Lin et al.

Code review is a crucial practice in software development. As code review nowadays is lightweight, various issues can be identified, and sometimes, they can be trivial. Research has investigated automated approaches to classify review comments to gauge the effectiveness of code reviews. However, previous studies have primarily relied on supervised machine learning, which requires extensive manual annotation to train the models effectively. To address this limitation, we explore the potential of using Large Language Models (LLMs) to classify code review comments. We assess the performance of LLMs to classify 17 categories of code review comments. Our results show that LLMs can classify code review comments, outperforming the state-of-the-art approach using a trained deep learning model. In particular, LLMs achieve better accuracy in classifying the five most useful categories, which the state-of-the-art approach struggles with due to low training examples. Rather than relying solely on a specific small training data distribution, our results show that LLMs provide balanced performance across high- and low-frequency categories. These results suggest that the LLMs could offer a scalable solution for code review analytics to improve the effectiveness of the code review process.

Chunhua Liu, Hong Yi Lin, Patanamon Thongtanunam

Language models have shown strong capabilities across a wide range of tasks in software engineering, such as code generation, yet they suffer from hallucinations. While hallucinations have been studied independently in natural language and code generation, their occurrence in tasks involving code changes which have a structurally complex and context-dependent format of code remains largely unexplored. This paper presents the first comprehensive analysis of hallucinations in two critical tasks involving code change to natural language generation: commit message generation and code review comment generation. We quantify the prevalence of hallucinations in recent language models and explore a range of metric-based approaches to automatically detect them. Our findings reveal that approximately 50\% of generated code reviews and 20\% of generated commit messages contain hallucinations. Whilst commonly used metrics are weak detectors on their own, combining multiple metrics substantially improves performance. Notably, model confidence and feature attribution metrics effectively contribute to hallucination detection, showing promise for inference-time detection.\footnote{All code and data will be released upon acceptance.

Chunhua Liu, Trevor Cohn, Lea Frermann

Humans use countless basic, shared facts about the world to efficiently navigate in their environment. This commonsense knowledge is rarely communicated explicitly, however, understanding how commonsense knowledge is represented in different paradigms is important for both deeper understanding of human cognition and for augmenting automatic reasoning systems. This paper presents an in-depth comparison of two large-scale resources of general knowledge: ConcpetNet, an engineered relational database, and SWOW a knowledge graph derived from crowd-sourced word associations. We examine the structure, overlap and differences between the two graphs, as well as the extent to which they encode situational commonsense knowledge. We finally show empirically that both resources improve downstream task performance on commonsense reasoning benchmarks over text-only baselines, suggesting that large-scale word association data, which have been obtained for several languages through crowd-sourcing, can be a valuable complement to curated knowledge graphs

Yan Zhao, Lu Liu, Chunhua Liu et al.

We introduce a new task named Story Ending Generation (SEG), whic-h aims at generating a coherent story ending from a sequence of story plot. Wepropose a framework consisting of a Generator and a Reward Manager for thistask. The Generator follows the pointer-generator network with coverage mech-anism to deal with out-of-vocabulary (OOV) and repetitive words. Moreover, amixed loss method is introduced to enable the Generator to produce story endingsof high semantic relevance with story plots. In the Reward Manager, the rewardis computed to fine-tune the Generator with policy-gradient reinforcement learn-ing (PGRL). We conduct experiments on the recently-introduced ROCStoriesCorpus. We evaluate our model in both automatic evaluation and human evalua-tion. Experimental results show that our model exceeds the sequence-to-sequencebaseline model by 15.75% and 13.57% in terms of CIDEr and consistency scorerespectively.

Chunhua Liu, Yan Zhao, Qingyi Si et al.

Commonsense Reading Comprehension (CRC) is a significantly challenging task, aiming at choosing the right answer for the question referring to a narrative passage, which may require commonsense knowledge inference. Most of the existing approaches only fuse the interaction information of choice, passage, and question in a simple combination manner from a \emph{union} perspective, which lacks the comparison information on a deeper level. Instead, we propose a Multi-Perspective Fusion Network (MPFN), extending the single fusion method with multiple perspectives by introducing the \emph{difference} and \emph{similarity} fusion\deleted{along with the \emph{union}}. More comprehensive and accurate information can be captured through the three types of fusion. We design several groups of experiments on MCScript dataset \cite{Ostermann:LREC18:MCScript} to evaluate the effectiveness of the three types of fusion respectively. From the experimental results, we can conclude that the difference fusion is comparable with union fusion, and the similarity fusion needs to be activated by the union fusion. The experimental result also shows that our MPFN model achieves the state-of-the-art with an accuracy of 83.52\% on the official test set.

Chunhua Liu, Haiou Zhang, Shan Jiang et al.

This paper proposes a Distilled-Exposition Enhanced Matching Network (DEMN) for story-cloze test, which is still a challenging task in story comprehension. We divide a complete story into three narrative segments: an \textit{exposition}, a \textit{climax}, and an \textit{ending}. The model consists of three modules: input module, matching module, and distillation module. The input module provides semantic representations for the three segments and then feeds them into the other two modules. The matching module collects interaction features between the ending and the climax. The distillation module distills the crucial semantic information in the exposition and infuses it into the matching module in two different ways. We evaluate our single and ensemble model on ROCStories Corpus \cite{Mostafazadeh2016ACA}, achieving an accuracy of 80.1\% and 81.2\% on the test set respectively. The experimental results demonstrate that our DEMN model achieves a state-of-the-art performance.

Chunhua Liu, Shan Jiang, Hainan Yu et al.

Natural Language Inference (NLI) is a fundamental and challenging task in Natural Language Processing (NLP). Most existing methods only apply one-pass inference process on a mixed matching feature, which is a concatenation of different matching features between a premise and a hypothesis. In this paper, we propose a new model called Multi-turn Inference Matching Network (MIMN) to perform multi-turn inference on different matching features. In each turn, the model focuses on one particular matching feature instead of the mixed matching feature. To enhance the interaction between different matching features, a memory component is employed to store the history inference information. The inference of each turn is performed on the current matching feature and the memory. We conduct experiments on three different NLI datasets. The experimental results show that our model outperforms or achieves the state-of-the-art performance on all the three datasets.