ARFBench - Presence: 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 Presence questions (whether the series shows an anomaly in the given time range), in percent; higher is better.

Metric: Accuracy (%) on the 111 Presence questions (is there an anomaly in the given time range, yes or no) 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
1Claude Opus 4.688.3#113
2Toto-1.0-QA-Experimental 32B (TSFM-VLM)84.7
3Qwen 3 VL 32B (post-trained)84.7
4Claude Sonnet 4.583.8#261
5Gemini 3 Pro82.9#145
6GPT-582#178
7GPT-5.4Medium81.1#135
8GPT-4.180.2#505
9Qwen 3 VL 8B80.2#917
10Qwen 3 VL 32B80.2#622
11GPT-4o79.3#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-presence · How It Works · Data refreshed daily, snapshot 2026-11-21.