Deep Learning (DL) models to analyze source code have shown immense promise during the past few years. More recently, self-supervised pre-training has gained traction for learning generic code representations valuable for many downstream SE tasks, such as clone and bug detection.
While previous work successfully learned from different code abstractions (e.g., token, AST, graph), we argue that it is also essential to factor in how developers code day-to-day for general-purpose representation learning. On the one hand, human developers tend to write repetitive programs referencing existing code snippets from the current codebase or online resources (e.g., Stack Overflow website) rather than implementing functions from scratch; such behaviors result in a vast number of code clones. In contrast, a deviant clone by mistake might trigger malicious program behaviors.
Thus, as a proxy to incorporate developers' coding behavior into the pre-training scheme, we propose to include code clones and their deviants. In particular, we propose CONCORD, a self-supervised, contrastive learning strategy to place benign clones closer in the representation space while moving deviants further apart. We show that CONCORD's clone-aware contrastive learning drastically reduces the need for expensive pre-training resources while improving the performance of downstream SE tasks. We also empirically demonstrate that CONCORD can improve existing pre-trained models to learn better representations that consequently become more efficient in both identifying semantically equivalent programs and differentiating buggy from non-buggy code.
more »
« less
This content will become publicly available on February 6, 2025
TRACED: Execution-aware Pre-training for Source Code
Most existing pre-trained language models for source code focus on learning the static code text, typically augmented with static code structures (abstract syntax tree, dependency graphs, etc.). However, program semantics will not be fully exposed before the real execution. Without an understanding of the program execution, statically pre-trained models fail to comprehensively capture the dynamic code properties, such as the branch coverage and the runtime variable values, and they are consequently less effective at code understanding tasks, such as retrieving semantic clones and detecting software vulnerabilities.
To close the gap between the static nature of language models and the dynamic characteristics of programs, we introduce TRACED, an execution-aware pre-training strategy for source code. Specifically, we pre-train code language models with a combination of source code, executable inputs, and corresponding execution traces. Our goal is to teach code models the complicated execution logic during the pre-training, enabling the model to statically estimate the dynamic code properties without repeatedly executing code during task-specific fine-tuning.
To illustrate the effectiveness of our proposed approach, we fine-tune and evaluate TRACED on three downstream tasks: static execution estimation, clone retrieval, and vulnerability detection. The empirical results show that TRACED relatively improves the statically pre-trained code models by 12.4% for complete execution path prediction and by 25.2% for runtime variable value predictions. TRACED also significantly outperforms statically pre-trained models in clone retrieval and vulnerability detection across four public benchmarks.
more »
« less
- NSF-PAR ID:
- 10523083
- Publisher / Repository:
- ACM
- Date Published:
- ISBN:
- 9798400702174
- Page Range / eLocation ID:
- 1 to 12
- Format(s):
- Medium: X
- Location:
- Lisbon Portugal
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Recent work learns contextual representations of source code by reconstructing tokens from their context. For downstream semantic understanding tasks like code clone detection, these representations should ideally capture program functionality. However, we show that the popular reconstruction-based RoBERTa model is sensitive to source code edits, even when the edits preserve semantics. We propose ContraCode: a contrastive pre-training task that learns code functionality, not form. ContraCode pre-trains a neural network to identify functionally similar variants of a program among many non-equivalent distractors. We scalably generate these variants using an automated source-to-source compiler as a form of data augmentation. Contrastive pre-training outperforms RoBERTa on an adversarial code clone detection benchmark by 39% AUROC. Surprisingly, improved adversarial robustness translates to better accuracy over natural code; ContraCode improves summarization and TypeScript type inference accuracy by 2 to 13 percentage points over competitive baselines. All source is available at https://github.com/parasj/contracode.more » « less
-
Vision-language (VL) pre-training has recently received considerable attention. However, most existing end-to-end pre-training approaches either only aim to tackle VL tasks such as image-text retrieval, visual question answering (VQA) and image captioning that test high-level understanding of images, or only target region-level understanding for tasks such as phrase grounding and object detection. We present FIBER (Fusion-In-the-Backbone-based transformER), a new VL model architecture that can seamlessly handle both these types of tasks. Instead of having dedicated transformer layers for fusion after the uni-modal backbones, FIBER pushes multimodal fusion deep into the model by inserting cross-attention into the image and text backbones to better capture multimodal interactions. In addition, unlike previous work that is either only pre-trained on image-text data or on fine-grained data with box-level annotations, we present a two-stage pre-training strategy that uses both these kinds of data efficiently: (i) coarse-grained pre-training based on image-text data; followed by (ii) fine-grained pre-training based on image-text-box data. We conduct comprehensive experiments on a wide range of VL tasks, ranging from VQA, image captioning, and retrieval, to phrase grounding, referring expression comprehension, and object detection. Using deep multimodal fusion coupled with the two-stage pre-training, FIBER provides consistent performance improvements over strong baselines across all tasks, often outperforming methods using magnitudes more data. Code is released at https://github.com/microsoft/FIBER.more » « less
-
Successful cross-language clone detection could enable researchers and developers to create robust language migration tools, facilitate learning additional programming languages once one is mastered, and promote reuse of code snippets over a broader codebase. How- ever, identifying cross-language clones presents special challenges to the clone detection problem. A lack of common underlying rep- resentation between arbitrary languages means detecting clones requires one of the following solutions: 1) a static analysis frame- work replicated across each targeted language with annotations matching language features across all languages, or 2) a dynamic analysis framework that detects clones based on runtime behavior. In this work, we demonstrate the feasibility of the latter solution, a dynamic analysis approach called SLACC for cross-language clone detection. Like prior clone detection techniques, we use input/out- put behavior to match clones, though we overcome limitations of prior work by amplifying the number of inputs and covering more data types; and as a result, achieve better clusters than prior at- tempts. Since clusters are generated based on input/output behav- ior, SLACC supports cross-language clone detection. As an added challenge, we target a static typed language, Java, and a dynamic typed language, Python. Compared to HitoshiIO, a recent clone de- tection tool for Java, SLACC retrieves 6 times as many clusters and has higher precision (86.7% vs. 30.7%). This is the first work to perform clone detection for dynamic typed languages (precision = 87.3%) and the first to perform clone detection across languages that lack a common underlying repre- sentation (precision = 94.1%). It provides a first step towards the larger goal of scalable language migration tools.more » « less
-
When software engineering researchers discuss "similar" code, we often mean code determined by static analysis to be textually, syntactically or structurally similar, known as code clones (looks alike). Ideally, we would like to also include code that is behaviorally or functionally similar, even if it looks completely different. The state of the art in detecting these behavioral clones focuses on checking the functional equivalence of the inputs and outputs of code fragments, regardless of its internal behavior (focusing only on input and output states). We argue that with an advance in dynamic code clone detection towards detecting behavioral clones (i.e., those with similar execution behavior), we can greatly increase the applications of behavioral clones as a whole for general program understanding tasks.more » « less
