--- dataset_info: features: - name: qid dtype: string - name: ground_truth_solution dtype: string - name: image_description dtype: string - name: test_script dtype: string - name: function_signature dtype: string - name: image dtype: image splits: - name: test num_bytes: 12840101 num_examples: 108 download_size: 12571814 dataset_size: 12840101 configs: - config_name: default data_files: - split: test path: data/test-* license: apache-2.0 task_categories: - image-to-text language: - en tags: - code pretty_name: humanevalv --- ## HumanEval-V: Evaluating Visual Understanding and Reasoning Abilities of LMMs Through Coding Tasks
📄 Paper • 🏠 Home Page • 💻 GitHub Repository • 🏆 Leaderboard • 🤗 Dataset Viewer
HumanEval-V includes 108 carefully crafted, entry-level Python coding tasks. LMMs are required to complete the code solution based on the provided visual context and a predefined Python function signature outlining the task requirements. Every task is equipped with meticulously handcrafted test cases for execution-based pass@k evaluation. ## Dataset Structure Each task in the dataset consists of the following fields: - **qid**: Unique identifier for each coding task (e.g., _q1_, with mutated versions like _q1-2_, _q1-3_). - **image**: A single image containing the essential visual context necessary to solve the task. - **function_signature**: Includes the problem description, necessary imports, and the function signature that the LMMs must complete. - **test_script**: Test cases used to validate the correctness of the generated code. - **ground_truth_solution**: Expert-crafted solutions provided for reference but not used during the evaluation process. - **image_description**: Human-labeled descriptions of the images, used for experimental analysis (not part of the benchmark evaluation). ## Prompt Format Each task is formatted with a clear instruction and provided function signature to guide the model in generating the code solution: ````markdown **Instructions:** You are an exceptionally intelligent coding assistant that consistently delivers accurate and reliable responses to user instructions. Please complete the function based on the provided image and code context. Return the complete solution, including the function signature, in a single response, formatted within a Python code block. **Code Context:** ```python {code_context} ``` ```` After the LMM generates a response, the code solution is extracted and validated using the following process: - Extraction of content within the code block. - Parsing the generated code to detect imports, class definitions, and functions using an Abstract Syntax Tree (AST) parser. - Concatenation of these components to form the final predicted solution, which is then tested for correctness. - Generated code solution is evaluated through an execution-based metric, specifically **pass@k**. ## Usage You can easily load the dataset using the Hugging Face `datasets` library. ```python from datasets import load_dataset humaneval_v = load_dataset("HumanEval-V/HumanEval-V-Benchmark", split="test") ``` ## Citation ```bibtex @article{zhang2024humanevalv, title={HumanEval-V: Evaluating Visual Understanding and Reasoning Abilities of Large Multimodal Models Through Coding Tasks}, author={Zhang, Fengji and Wu, Linquan and Bai, Huiyu and Lin, Guancheng and Li, Xiao and Yu, Xiao and Wang, Yue and Chen, Bei and Keung, Jacky}, journal={arXiv preprint arXiv:2410.12381}, year={2024}, } ```