ARFBench - Correlation: leaderboard

Time series question answering on software incident data in ARFBench (2026, Datadog): multiple-choice questions over multivariate production telemetry from real incidents, with the series shown as chart images. Accuracy on the Correlation questions (whether anomalies in two series are correlated), in percent; higher is better.

Metric: Accuracy (%) on the 170 Correlation questions (whether anomalies in two time series are correlated) of ARFBench, for the time series rendered as chart images (Matplotlib plots, three images for paired questions), with the series description and channel names as text; few-shot multiple-choice prompting with shuffled answer choices, temperature 0.05 (default for models without the setting), reasoning models at medium effort with 2,000 output tokens, blank or invalid answers counted wrong; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 Pro71.2#145
2Toto-1.0-QA-Experimental 32B (TSFM-VLM)68.8
3Qwen 3 VL 32B (post-trained)67.6
4Claude Opus 4.665.9#113
5GPT-563.5#178
6GPT-5.4Medium61.8#135
7Qwen 3 VL 32B61.8#622
8Claude Sonnet 4.558.8#261
9Qwen 3 VL 8B57.1#917
10GPT-4.145.9#505
11GPT-4o45.3#751

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=arfbench-correlation · How It Works · Data refreshed daily, snapshot 2026-11-21.