… and dataflow) into the sequential model pre-training. To further explore the … SYNCOBERT,
a Syntax-guided multi-modal contrastive pre-training approach for better Code representations…

… Compared to the previous state-of-theart SynCoBERT, CODE-MVP also performs better …
indicates that the code representation learned by CODE-MVP preserves more code semantics. …

… , several pre-training methods based on contrastive learning … studies, our model utilizes
multimodal contrastive learning by … SyncoBERT [13] aims to leverage the complementary …

… pretraining methods based on contrastive learning such as Corder [3], Code-MVP [35],
CodeRetriver [18], SynCoBERT [… code for joint encoding to obtain the code representation vector. …

… multi-modal contents to learn representation of code fragment with … data flow to enhance
code representation, and SYNCOBERT that … Syncobert: Syntax-guided multi-modal contrastive …

… an important way to learn code semantics of source code through pre-training, and the pre-…
of codes, SYNCoBERT [9] combines AST edge prediction and multimodal contrastive learning…

… code representations. We propose two-phase pre-training to comprehensively capture code
… extended to multi-modal tasks and perform well by loading a multi-modal pre-trained model. …

… Comment and code pairs can also be viewed as multi-modal … the loss function in the
pre-training process of SynCoBERT, C3P, … Jiang, “Syncobert: Syntax-guided multi-modal …

… time in the hope of turning it into a representation that makes it possible for the computer to
more … new pre-training tasks to effectively use this additional information. SynCoBERT is a …

… Specifically, in order to obtain better representations of code and query, we … SynCoBERT
proposed by Wang et al. [21], adopts a multi-modal contrastive learning objective for pre-training…