Interaction2Code: Benchmark for MLLM-Based Interactive Webpage Generation

Interaction2Code is the first systematic benchmark for Multimodal Large Language Models generating interactive webpages. It encompasses 127 unique webpages and 374 distinct interactions across 15 webpage types and 31 interaction categories. The dataset was created by whale99 and last updated on 2026-05-25.

Use Cases

• Benchmarking MLLM performance on interactive UI-to-code tasks based on the described 374 distinct interactions.
• Training models to generate functional webpage code based on the 15 webpage types mentioned.
• Analyzing the capabilities of AI models in handling complex web interactions across the 31 defined categories.

Strengths

• Benchmarks 127 unique webpages, providing a substantial testbed.
• Covers 374 distinct interactions, suggesting a focus on complex UI behavior.
• Spans 15 webpage types and 31 interaction categories, indicating breadth.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update is dated 2026-05-25.

Provenance

Source

whale99 via Hugging Face

Freshness

Last updated 2026-05-25 05:54:48