Papers with CodeBERT
Source Code is a Graph, Not a Sequence: A Cross-Lingual Perspective on Code Clone Detection (2024.naacl-srw)
Copied to clipboard
| Challenge: | Existing methods for code clone detection are fast and scalable, but they may fail to detect clones that have different syntax or structure. |
| Approach: | They propose to use graph-based methods for code clone detection instead of sequence-based models. |
| Outcome: | The proposed graph-based methods outperforms sequence-based models on cross-lingual code clones. |
Tighter Clusters, Safer Code? Improving Vulnerability Detection with Enhanced Contrastive Loss (2025.naacl-srw)
Copied to clipboard
| Challenge: | a recent study shows that supervised contrastive learning (SCL) improves embedding separation but struggles with intra-class clustering . code vulnerability detection is a cornerstone of software security, especially as the world undergoes rapid digitization. |
| Approach: | They propose an extension of SCL with a distance-based regularization term that tightens intra-class clustering while maintaining inter-class separation. |
| Outcome: | The proposed method improves F1 score on CodeBERT and GraphCodeBERT with BCE, BCE + SCL, and BCE+ CESCL. |
CodeBERT: A Pre-Trained Model for Programming and Natural Languages (2020.findings-emnlp)
Copied to clipboard
Zhangyin Feng, Daya Guo, Duyu Tang, Nan Duan, Xiaocheng Feng, Ming Gong, Linjun Shou, Bing Qin, Ting Liu, Daxin Jiang, Ming Zhou
| Challenge: | Large pre-trained models have improved performance on a variety of natural language processing tasks. |
| Approach: | They develop a bimodal pre-trained model for programming language (PL) and natural language (NL) it incorporates a hybrid objective function that detects replaced tokens from generators. |
| Outcome: | The proposed model performs better on two NL-PL applications by fine-tuning model parameters. |
HierarchyNet: Learning to Summarize Source Code with Heterogeneous Representations (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing code summarization approaches ignore the interplay of dependencies among program elements and code hierarchy. |
| Approach: | They propose a code summarization approach utilizing Heterogeneous Code Representations (HCRs) and HierarchyNet. |
| Outcome: | The proposed method improves on existing models and pre-trained models. |
The Vault: A Comprehensive Multilingual Dataset for Advancing Code Understanding and Generation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Open-source dataset of code-text pairs for training large language models to understand code is outperforms other datasets for code generation and understanding tasks. |
| Approach: | They propose to extract high-quality code-text pairs from a dataset of 43 million pairs . they use rules and deep learning to ensure that the code-sampled samples contain high-quality pairs a . |
| Outcome: | The Vault dataset outperforms existing models on common coding tasks . authors hope the results will propel AI research and software development forward . |
DIP: Dead code Insertion based Black-box Attack for Programming Language Model (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods to attack natural language models are difficult to apply due to the requirements. |
| Approach: | They propose a black-box attack method that generates adversarial examples using dead code insertion. |
| Outcome: | The proposed method outperforms the state-of-the-art black-box attack in both attack efficiency and attack quality on 9 victim downstream-task large code models. |
CoSQA: 20,000+ Web Queries for Code Search and Question Answering (2021.acl-long)
Copied to clipboard
| Challenge: | Using deep neural networks to find codes is difficult . we present a dataset that includes 20,604 labels for natural language queries and codes . |
| Approach: | They introduce a contrastive learning method to enhance text-code matching . they find that CoSQA improves the accuracy of code question answering by 5.1% . |
| Outcome: | The proposed method improves the accuracy of code question answering by 5.1% and improves by 10.5% on a CodeBERT model. |
DCE-LLM: Dead Code Elimination with Large Language Models (2025.naacl-long)
Copied to clipboard
| Challenge: | Dead code can obscure logical errors and be exploited for obfuscation in malware. |
| Approach: | They propose a framework for automated dead code elimination using a codeBERT model with an attribution-based line selector. |
| Outcome: | Experimental results show that DCE-LLM outperforms existing tools for dead code elimination . dead code can obscure logical errors and be exploited for obfuscation in malware . |
A Black-Box Attack on Code Models via Representation Nearest Neighbor Search (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for generating adversarial code examples face challenges such as limted availability of substitute variables and the creation of adversarials with noticeable perturbations. |
| Approach: | They propose a search seed based on historical attacks to find adversarial substitutes . they employ a pre-trained variable name encoder to map the search seed to a continuous vector space . |
| Outcome: | The proposed approach outperforms baseline methods in terms of ASR and QT. |
Better Language Models of Code through Self-Improvement (2023.findings-acl)
Copied to clipboard
| Challenge: | Pre-trained language models for code (PLMCs) are pre-taught on large datasets with multi-modal objectives, but fine-tuning them requires extensive supervision and is limited by the size of the dataset provided. |
| Approach: | They propose a data augmentation framework that utilizes knowledge from the pre-training and fine-tuning stage to augment training data, which is then used for the next step. |
| Outcome: | The proposed framework significantly improves pre-trained language models’ performance in sequence-generation tasks, such as code summarization and code generation in the CodeXGLUE benchmark. |
Tackling Long Code Search with Splitting, Encoding, and Aggregating (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing pretraining models take the first 256 tokens of code snippets by default, limiting the input length to 512. |
| Approach: | They propose a baseline SEA model which splits long code into code blocks and aggregates them to obtain a comprehensive long code representation. |
| Outcome: | The proposed model can model long code without changing their internal structure and re-pretraining. |