ArtifactBench - Attribute Prediction: leaderboard
Predicting reported evaluation results on ArtifactBench (Yu et al., 2026), a graph of Hugging Face models, datasets and evaluation records: for each model-dataset pair in the test split, the LLM predicts the reported score, normalized to 0-1, from the model and dataset metadata alone. Mean absolute error; lower is better.
Metric: Mean absolute error (0-1; predicted vs reported score on model-dataset evaluation edges of a Hugging Face artifact graph (transductive test split), targets are the reported metric values normalized to 0-1; the LLM reads the model and dataset metadata without graph context). Source: arxiv.org. Saturation forecast: Around February 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 0.14 | #172 |
| 2 | Qwen 3 8B | 0.2 | #1838 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=artifactbench-attribute-prediction · How It Works · Data refreshed daily, snapshot 2026-11-12.