BatCoder: Self-Supervised Bidirectional Code-Documentation Learning via Back-Translation
This addresses the scarcity of high-quality code-documentation pairs, especially for niche programming languages, by enabling training with only code, though it is incremental as it builds on existing self-supervised and back-translation techniques.
The paper tackles the problem of training LLMs for code-related tasks without costly code-documentation pairs by introducing BatCoder, a self-supervised framework that uses back-translation to jointly optimize code generation and documentation production, achieving 83.5% and 81.0% pass@1 on HumanEval and MBPP benchmarks.
Training LLMs for code-related tasks typically depends on high-quality code-documentation pairs, which are costly to curate and often scarce for niche programming languages. We introduce BatCoder, a self-supervised reinforcement learning framework designed to jointly optimize code generation and documentation production. BatCoder employs a back-translation strategy: a documentation is first generated from code, and then the generated documentation is used to reconstruct the original code. The semantic similarity between the original and reconstructed code serves as an implicit reward, enabling reinforcement learning to improve the model's performance both in generating code from documentation and vice versa. This approach allows models to be trained using only code, substantially increasing the available training examples. Evaluated on HumanEval and MBPP with a 7B model, BatCoder achieved 83.5% and 81.0% pass@1, outperforming strong open-source baselines. Moreover, the framework demonstrates consistent scaling with respect to both training corpus size and model capacity.