ReactBench (Hallucination) - Alteration Tracing (CoT): leaderboard
Accuracy in percent on ReactBench (2026) questions about before-and-after image pairs, asking what changed, with a step-by-step reasoning prompt; an incorrect answer counts as a hallucination. Higher is better.
Metric: Accuracy (%) on Alteration Tracing (10,956 questions on 913 before-and-after image pairs) with a step-by-step reasoning prompt, hallucination-inducing questions (multiple choice, short answer, yes/no) about images edited by Qwen-Image-Edit-2511; an incorrect answer counts as a hallucination; temperature 0.2, at most 1,024 tokens, LLM answer matching; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 32B (Thinking) | 78.7 | #573 |
| 2 | Qwen 3.5 27B | 73.7 | #375 |
| 3 | ReactBench MiMo-VL-7B (checkpoint unspecified) | 73.5 | |
| 4 | Qwen 3.5 9B | 63.9 | #828 |
| 5 | Qwen 3 VL 32B Instruct | 63.5 | #606 |
| 6 | InternVL3-8B | 60.7 | #1782 |
| 7 | Qwen 3 VL 4B | 53.4 | #1151 |
| 8 | Qwen 2.5 VL 72B Instruct | 50.5 | #841 |
| 9 | Qwen 2.5 VL 7B Instruct | 41.9 | #1902 |
| 10 | InternVL2.5-8B | 38.3 | #2370 |
| 11 | InternVL2.5-26B | 37.7 | #2156 |
| 12 | ReactBench LLaVA-OV-7B (checkpoint unspecified) | 34.7 | |
| 13 | ReactBench LLaVA-v1.6-7B (checkpoint unspecified) | 34 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=reactbench-hallucination-alteration-tracing-cot · How It Works · Data refreshed daily, snapshot 2026-11-21.