Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2510.14115 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918161601789952 |
|---|---|
| author | Bauerfeind, Philipp Salarpour, Amir Fernandez, David MohajerAnsari, Pedram Reschke, Johannes Pesé, Mert D. |
| author_facet | Bauerfeind, Philipp Salarpour, Amir Fernandez, David MohajerAnsari, Pedram Reschke, Johannes Pesé, Mert D. |
| contents | Scenario simulation is central to testing autonomous driving systems. Scenic, a domain-specific language (DSL) for CARLA, enables precise and reproducible scenarios, but NL-to-Scenic generation with large language models (LLMs) suffers from scarce data, limited reproducibility, and inconsistent metrics. We introduce NL2Scenic, an open dataset and framework with 146 NL/Scenic pairs, a difficulty-stratified 30-case test split, an Example Retriever, and 14 prompting variants (ZS, FS, CoT, SP, MoT). We evaluate 13 models: four proprietary (GPT-4o, GPT-5, Claude-Sonnet-4, Gemini-2.5-pro) and nine open-source code models (Qwen2.5Coder 0.5B-32B; CodeLlama 7B/13B/34B), using text metrics (BLEU, ChrF, EDIT-SIM, CrystalBLEU) and execution metrics (compilation and generation), and compare them with an expert study (n=11). EDIT-SIM correlates best with human judgments; we also propose EDIT-COMP (F1 of EDIT-SIM and compilation) as a robust dataset-level proxy that improves ranking fidelity. GPT-4o performs best overall, while Qwen2.5Coder-14B reaches about 88 percent of its expert score on local hardware. Retrieval-augmented prompting, Few-Shot with Example Retriever (FSER), consistently boosts smaller models, and scaling shows diminishing returns beyond mid-size, with Qwen2.5Coder outperforming CodeLlama at comparable scales. NL2Scenic and EDIT-COMP offer a standardized, reproducible basis for evaluating Scenic code generation and indicate that mid-size open-source models are practical, cost-effective options for autonomous-driving scenario programming. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_14115 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | David vs. Goliath: A comparative study of different-sized LLMs for code generation in the domain of automotive scenario generation Bauerfeind, Philipp Salarpour, Amir Fernandez, David MohajerAnsari, Pedram Reschke, Johannes Pesé, Mert D. Software Engineering Machine Learning Scenario simulation is central to testing autonomous driving systems. Scenic, a domain-specific language (DSL) for CARLA, enables precise and reproducible scenarios, but NL-to-Scenic generation with large language models (LLMs) suffers from scarce data, limited reproducibility, and inconsistent metrics. We introduce NL2Scenic, an open dataset and framework with 146 NL/Scenic pairs, a difficulty-stratified 30-case test split, an Example Retriever, and 14 prompting variants (ZS, FS, CoT, SP, MoT). We evaluate 13 models: four proprietary (GPT-4o, GPT-5, Claude-Sonnet-4, Gemini-2.5-pro) and nine open-source code models (Qwen2.5Coder 0.5B-32B; CodeLlama 7B/13B/34B), using text metrics (BLEU, ChrF, EDIT-SIM, CrystalBLEU) and execution metrics (compilation and generation), and compare them with an expert study (n=11). EDIT-SIM correlates best with human judgments; we also propose EDIT-COMP (F1 of EDIT-SIM and compilation) as a robust dataset-level proxy that improves ranking fidelity. GPT-4o performs best overall, while Qwen2.5Coder-14B reaches about 88 percent of its expert score on local hardware. Retrieval-augmented prompting, Few-Shot with Example Retriever (FSER), consistently boosts smaller models, and scaling shows diminishing returns beyond mid-size, with Qwen2.5Coder outperforming CodeLlama at comparable scales. NL2Scenic and EDIT-COMP offer a standardized, reproducible basis for evaluating Scenic code generation and indicate that mid-size open-source models are practical, cost-effective options for autonomous-driving scenario programming. |
| title | David vs. Goliath: A comparative study of different-sized LLMs for code generation in the domain of automotive scenario generation |
| topic | Software Engineering Machine Learning |
| url | https://arxiv.org/abs/2510.14115 |