ScratchEval : A Multimodal Evaluation Framework for LLMs in Block-Based Programming

TL;DR

ScratchEval is a comprehensive benchmark designed to evaluate the ability of large language models to understand, debug, and repair Scratch programs, addressing the unique challenges of block-based, event-driven programming.

Contribution

This work introduces the first executable benchmark for LLMs on Scratch, including a detailed evaluation protocol and curated dataset focusing on complex, real-world projects.

Findings

LLMs show limited performance on Scratch-specific tasks

Fine-tuning improves repair accuracy but still faces challenges

Benchmark enables systematic assessment of model understanding and repair quality

Abstract

LLMs have achieved strong performance on text-based programming tasks, yet they remain unreliable for block-based languages such as Scratch. Scratch programs exhibit deeply nested, non-linear structures, event-driven concurrency across multiple sprites, and tight coupling between code and multimedia assets, properties that differ fundamentally from textual code. As a result, LLMs often misinterpret Scratch semantics and generate large, invasive edits that are syntactically valid but semantically incorrect when repairing buggy programs. We introduce ScratchEval, the first executable benchmark designed to evaluate LLM-based repair for Scratch programs, covering program understanding, debugging, analysis, and repair. The benchmark contains 100 curated Scratch projects from the public repository, selected for structural and semantic complexity. Each project is paired with executable test suites, bug descriptions with corresponding fixes, block-level edit constraints defining minimal semantically correct repairs, and required multimedia assets. The benchmark is constructed through a human-in-the-loop pipeline combining automated project mining with expert validation of trigger-outcome semantics and representative bug patterns, with emphasis on event ordering, concurrency, and state management. To enable rigorous and reproducible evaluation, we propose a three-layer executable protocol measuring functional correctness via VM-level execution, repair quality using block-level edit distance and behavioral trajectory comparisons, and explanation quality via structured rubrics assessing alignment between model reasoning and generated patches. Using ScratchEval, we study domain-specific fine-tuning, training data effectiveness, and model generalization to unseen bug types. ScratchEval provides a reproducible foundation for evaluating and post-training LLMs on block-based programming tasks.