VLM-DeflectionBench - Oracle Accuracy: leaderboard
VLM-DeflectionBench (2026) tests knowledge-based visual question answering where the right response may be to decline when retrieved evidence is missing or misleading; a GPT-4o judge grades each answer correct, deflected or hallucinated. This board is accuracy in percent with only gold evidence; higher is better.
Metric: Accuracy (%) of knowledge-based visual question answering in the Oracle retrieval scenario (gold evidence only), judged by GPT-4o with the SimpleQA protocol; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 19 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 73.1 | #185 |
| 2 | Ovis2-34B | 66.5 | #1061 |
| 3 | MiniCPM-V-4.5 | 65.3 | #1451 |
| 4 | Ovis2-16B | 64.4 | #1608 |
| 5 | InternVL3-38B | 64.2 | #948 |
| 6 | Keye-VL-1.5-8B | 63.4 | #1729 |
| 7 | Pixtral-12B | 62.7 | #2611 |
| 8 | VL-Rethinker-72B | 62.6 | #1252 |
| 9 | Aria | 62.3 | #2190 |
| 10 | InternVL3-9B | 61.2 | #1746 |
| 11 | Qwen 2.5 VL 32B Instruct | 61 | #1065 |
| 12 | Gemini 2.5 Pro | 59.8 | #280 |
| 13 | Gemma 3 27B | 59.5 | #1672 |
| 14 | Gemini 2.5 Flash | 58.8 | #466 |
| 15 | GLM-4.1V-9B (Thinking) | 55.2 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=vlm-deflectionbench-oracle-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-22.