# Output Format Biases in the Evaluation of Large Language Models for Code Translation

arXiv: https://arxiv.org/abs/2403.17214 · HTML: https://mlsanity.com/paper/2403.17214/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 62 references that are in the map.
- First posted on arXiv 25 Mar 2024.

## Builds on (most cited first, 25 of 25)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [HuggingFace's Transformers: State-of-the-art Natural Language Processing](https://mlsanity.com/paper/1910.03771/) (2019)
- [Code Llama: Open Foundation Models for Code](https://mlsanity.com/paper/2308.12950/) (2023)
- [CodeBERT: A Pre-Trained Model for Programming and Natural Languages](https://mlsanity.com/paper/2002.08155/) (2020)
- [CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation](https://mlsanity.com/paper/2109.00859/) (2021)
- [CodeGen: An Open Large Language Model for Code with Multi-Turn Program Synthesis](https://mlsanity.com/paper/2203.13474/) (2022)
- [StarCoder: may the source be with you!](https://mlsanity.com/paper/2305.06161/) (2023)
- [CodeXGLUE: A Machine Learning Benchmark Dataset for Code Understanding and Generation](https://mlsanity.com/paper/2102.04664/) (2021)
- [WizardCoder: Empowering Code Large Language Models with Evol-Instruct](https://mlsanity.com/paper/2306.08568/) (2023)
- [CodeBLEU: a Method for Automatic Evaluation of Code Synthesis](https://mlsanity.com/paper/2009.10297/) (2020)
- [CodeGeeX: A Pre-Trained Model for Code Generation with Multilingual Benchmarking on HumanEval-X](https://mlsanity.com/paper/2303.17568/) (2023)
- [Unsupervised Translation of Programming Languages](https://mlsanity.com/paper/2006.03511/) (2020)
- [CodeNet: A Large-Scale AI for Code Dataset for Learning a Diversity of Coding Tasks](https://mlsanity.com/paper/2105.12655/) (2021)
- [Large Language Models for Software Engineering: Survey and Open Problems](https://mlsanity.com/paper/2310.03533/) (2023)
- [CodeBERTScore: Evaluating Code Generation with Pretrained Models of Code](https://mlsanity.com/paper/2302.05527/) (2023)
- [Tree-to-tree Neural Networks for Program Translation](https://mlsanity.com/paper/1802.03691/) (2018)
- [Leveraging Automated Unit Tests for Unsupervised Code Translation](https://mlsanity.com/paper/2110.06773/) (2021)
- [Large Language Models are Few-Shot Summarizers: Multi-Intent Comment Generation via In-Context Learning](https://mlsanity.com/paper/2304.11384/) (2023)
- [DOBF: A Deobfuscation Pre-Training Objective for Programming Languages](https://mlsanity.com/paper/2102.07492/) (2021)
- [XLCoST: A Benchmark Dataset for Cross-lingual Code Intelligence](https://mlsanity.com/paper/2206.08474/) (2022)
- [Perfection Not Required? Human-AI Partnerships in Code Translation](https://mlsanity.com/paper/2104.03820/) (2021)
- [Prompting Is All You Need: Automated Android Bug Replay with Large Language Models](https://mlsanity.com/paper/2306.01987/) (2023)
- [Nuances are the Key: Unlocking ChatGPT to Find Failure-Inducing Tests with Differential Prompting](https://mlsanity.com/paper/2304.11686/) (2023)
- [On the Evaluation of Neural Code Translation: Taxonomy and Benchmark](https://mlsanity.com/paper/2308.08961/) (2023)

## Built on by (most cited first, 9 of 9)

- [LLMs Are Biased Towards Output Formats! Systematically Evaluating and Mitigating Output Format Bias of LLMs](https://mlsanity.com/paper/2408.08656/) (2024)
- [InterTrans: Leveraging Transitive Intermediate Translations to Enhance LLM-based Code Translation](https://mlsanity.com/paper/2411.01063/) (2024)
- [Evaluation of LLMs for mathematical problem solving](https://mlsanity.com/paper/2506.00309/) (2025)
- [Function-to-Style Guidance of LLMs for Code Translation](https://mlsanity.com/paper/2507.11083/) (2025)
- [Can LLMs Extract Frame-Semantic Arguments?](https://mlsanity.com/paper/2502.12516/) (2025)
- [Post-Incorporating Code Structural Knowledge into Pretrained Models via ICL for Code Translation](https://mlsanity.com/paper/2503.22776/) (2025)
- [RepoTransAgent: Multi-Agent LLM Framework for Repository-Aware Code Translation](https://mlsanity.com/paper/2508.17720/) (2025)
- [Towards Conversational Development Environments: Using Theory-of-Mind and Multi-Agent Architectures for Requirements Refinement](https://mlsanity.com/paper/2505.20973/) (2025)
- [The New Compiler Stack: A Survey on the Synergy of LLMs and Compilers](https://mlsanity.com/paper/2601.02045/) (2026)

## Abstract (arXiv)

Code translation between programming languages (PLs) is a critical task in software engineering, facilitating the modernization of legacy systems, ensuring cross-platform compatibility, and enhancing software performance. Most existing studies instruct LLMs to perform code translation and evaluate their performance by either running the generated outputs through test suites or comparing them to reference outputs (ground truth). These outputs, however, may contain not only executable source code but also additional non-code elements, such as natural language explanations or formatting tokens. We refer to the combination of source code and non-code elements as the output format. It is crucial to understand and address variations in output format, as non-code elements can interfere with evaluation metrics, resulting in biased assessments of model performance and comparisons. We conduct an empirical analysis of the outputs from eleven instruct-tuned open-source LLMs, across five PLs: C, C++, Go, Java, and Python. The results show that between 26.4% and 73.7% of outputs produced by our evaluated LLMs necessitate post-processing. To mitigate output format bias, we propose a strategic combination of prompt engineering and regular expressions that effectively extracts source code from mixed-format outputs, enabling the eleven open-source models to achieve an average Code Extraction Success Rate (CSR) of 92.73%. Our empirical study confirms that output format bias affects widely used execution-based metrics, i.e., Computational Accuracy (CA), and text-based metrics, i.e., BLEU, CodeBLEU and CrystalBLEU. Additionally, we test five closed-source LLMs and observe that they also generate varying distributions of output formats, which could lead to output format biases. Our results highlight the need to mitigate the output format bias to enable reliable evaluations in LLMs for code translation.