Pretrained language models have been shown to be effective in many software-related generation tasks; however, they are not well-suited for editing tasks as they are not designed to reason about edits. To address this, we propose a novel pretraining objective which explicitly models edits and use it to build CoditT5, a large language model for software-related editing tasks that is pretrained on large amounts of source code and natural language comments. We fine-tune it on various downstream editing tasks, including comment updating, bug fixing, and automated code review. By outperforming standard generation-based models, we demonstrate the generalizability of our approach and its suitability for editing tasks. We also show how a standard generation model and our edit-based model can complement one another through simple reranking strategies, with which we achieve state-of-the-art performance for the three downstream editing tasks.

/pdf/coditt5-pretraining-for-source-code-and-natural-language-1v0zscvi.pdf

CoditT5: Pretraining for Source Code and Natural Language Editing

Recent years have seen the successful application of deep learning to software engineering (SE). In particular, the development and use of pre-trained models of source code has enabled state-of-the-art results to be achieved on a wide variety of SE tasks. This paper provides an overview of this rapidly advancing field of research and reflects on future research directions.

Deep Learning Meets Software Engineering: A Survey on Pre-Trained Models of Source Code

Large language models (LLMs) pretrained on vast source code have achieved prominent progress in code intelligence. However, existing code LLMs have two main limitations in terms of architecture and pretraining tasks. First, they often adopt a specific architecture (encoder-only or decoder-only) or rely on a unified encoder-decoder network for different downstream tasks. The former paradigm is limited by inflexibility in applications while in the latter, the model is treated as a single system for all tasks, leading to suboptimal performance on a subset of tasks. Secondly, they often employ a limited set of pretraining objectives which might not be relevant to some downstream tasks and hence result in substantial performance degrade. To address these limitations, we propose ``CodeT5+'', a family of encoder-decoder LLMs for code in which component modules can be flexibly combined to suit a wide range of downstream code tasks. Such flexibility is enabled by our proposed mixture of pretraining objectives to mitigate the pretrain-finetune discrepancy. These objectives cover span denoising, contrastive learning, text-code matching, and causal LM pretraining tasks, on both unimodal and bimodal multilingual code corpora. Furthermore, we propose to initialize CodeT5+ with frozen off-the-shelf LLMs without training from scratch to efficiently scale up our models, and explore instruction-tuning to align with natural language instructions. We extensively evaluate CodeT5+ on over 20 code-related benchmarks in different settings, including zero-shot, finetuning, and instruction-tuning. We observe state-of-the-art (SoTA) model performance on various code-related tasks, such as code generation and completion, math programming, and text-to-code retrieval tasks. Particularly, our instruction-tuned CodeT5+ 16B achieves new SoTA results on HumanEval code generation task against other open code LLMs.

CodeT5+: Open Code Large Language Models for Code Understanding and Generation

While a large number of pre-trained models of source code have been successfully developed and applied to a variety of software engineering (SE) tasks in recent years, our understanding of these pre-trained models is arguably fairly limited. With the goal of advancing our understanding of these models, we perform the first systematic empirical comparison of 19 recently-developed pre-trained models of source code on 13 SE tasks. To gain additional insights into these models, we adopt a recently-developed 4-dimensional categorization of pre-trained models, and subsequently investigate whether there are correlations between different categories of pre-trained models and their performances on different SE tasks.

An Empirical Comparison of Pre-Trained Models of Source Code

The utilization of programming language (PL) models, pre-trained on large-scale code corpora, as a means of automating software engineering processes has demonstrated considerable potential in streamlining various code generation tasks such as code completion, code translation, and program synthesis. However, current approaches mainly rely on supervised fine-tuning objectives borrowed from text generation, neglecting unique sequence-level characteristics of code, including but not limited to compilability as well as syntactic and functional correctness. To address this limitation, we propose PPOCoder, a new framework for code generation that synergistically combines pre-trained PL models with Proximal Policy Optimization (PPO) which is a widely used deep reinforcement learning technique. By utilizing non-differentiable feedback from code execution and structure alignment, PPOCoder seamlessly integrates external code-specific knowledge into the model optimization process. It's important to note that PPOCoder is a task-agnostic and model-agnostic framework that can be used across different code generation tasks and PLs. Extensive experiments on three code generation tasks demonstrate the effectiveness of our proposed approach compared to SOTA methods, achieving significant improvements in compilation success rates and functional correctness across different PLs.

Execution-based Code Generation using Deep Reinforcement Learning

Recent years have seen the successful application of large pretrained models to code representation learning, resulting in substantial improvements on many code-related downstream tasks. But there are issues surrounding their application to SE tasks. First, the majority of the pre-trained models focus on pre-training only the encoder of the Transformer. For generation tasks that are addressed using models with the encoder-decoder architecture, however, there is no reason why the decoder should be left out during pre-training. Second, many existing pre-trained models, including state-of-the-art models such as T5-learning, simply reuse the pretraining tasks designed for natural languages. Moreover, to learn the natural language description of source code needed eventually for code-related tasks such as code summarization, existing pretraining tasks require a bilingual corpus composed of source code and the associated natural language description, which severely limits the amount of data for pre-training. To this end, we propose SPT-Code, a sequence-to-sequence pre-trained model for source code. In order to pre-train SPT-Code in a sequence-to-sequence manner and address the aforementioned weaknesses associated with existing pre-training tasks, we introduce three pre-training tasks that are specifically designed to enable SPT-Code to learn knowledge of source code, the corresponding code structure, as well as a natural language description of the code without relying on any bilingual corpus, and eventually exploit these three sources of information when it is applied to downstream tasks. Experimental results demonstrate that SPT-Code achieves state-of-the-art performance on five code-related downstream tasks after fine-tuning.

SPT-Code: Sequence-to-Sequence Pre-Training for Learning Source Code Representations

Automatic legal judgment prediction (LJP) has recently received increasing attention in the natural language processing community in part because of its practical values as well as the associated research challenges. We present an overview of the major milestones made in LJP research covering multiple jurisdictions and multiple languages, and conclude with promising future research directions.

/pdf/legal-judgment-prediction-a-survey-of-the-state-of-the-art-29dxytu4.pdf

Legal Judgment Prediction: A Survey of the State of the Art

Since 2009, the deep learning revolution, which was triggered by the introduction of ImageNet, has stimulated the synergy between Software Engineering (SE) and Machine Learning (ML)/Deep Learning (DL). Meanwhile, critical reviews have emerged that suggest that ML/DL should be used cautiously. To improve the applicability and generalizability of ML/DL-related SE studies, we conducted a 12-year Systematic Literature Review (SLR) on 1,428 ML/DL-related SE papers published between 2009 and 2020. Our trend analysis demonstrated the impacts that ML/DL brought to SE. We examined the complexity of applying ML/DL solutions to SE problems and how such complexity led to issues concerning the reproducibility and replicability of ML/DL studies in SE. Specifically, we investigated how ML and DL differ in data preprocessing, model training, and evaluation when applied to SE tasks, and what details need to be provided to ensure that a study can be reproduced or replicated. By categorizing the rationales behind the selection of ML/DL techniques into five themes, we analyzed how model performance, robustness, interpretability, complexity, and data simplicity affected the choices of ML/DL models.

Vincent Ng

Papers

SPT-Code: Sequence-to-Sequence Pre-Training for Learning Source Code Representations

Deep Learning Meets Software Engineering: A Survey on Pre-Trained Models of Source Code

An Empirical Comparison of Pre-Trained Models of Source Code

Legal Judgment Prediction: A Survey of the State of the Art

Machine/Deep Learning for Software Engineering: A Systematic Literature Review