CiteVQA - Answer Correctness: leaderboard

CiteVQA (2026) asks multimodal models to answer questions over long PDFs, 1,897 questions on 711 documents in seven domains, and to cite the exact page elements that support each answer as bounding boxes. This board is answer correctness alone, judged on a 0 to 5 scale and multiplied by 20, averaged over all questions; higher is better.

Metric: Answer correctness against the reference answer, judged on a 0-5 scale and multiplied by 20 (0-100), over all 1,897 questions, the model answers from the document pages and cites element-level bounding boxes, unified prompt, temperature 1.0, at most 4,096 output tokens, Qwen3-VL-235B-A22B judge; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 20 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.4xHigh87.1#135
2Gemini 3.1 Pro (Preview)High86.1#99
3Qwen 3.6 Plus85.9#238
4Gemini 3 Flash (Preview)High84.5#152
5Gemini 2.5 ProThinking, 32K82.2#276
6Seed 2.0 Pro81.3#222
7Qwen 3.5 397B A17B76.5#264
8Qwen 3.5 35B A3B76.4#492
9Qwen 3.5 27B75.6#375
10Kimi K2.574.3#266
11Qwen 3.5 122B A10B73.6#346
12Qwen 3 VL 32B Instruct72.3#606
13Qwen 3 VL 235B A22B Instruct72.3#532
14GPT-5.2xHigh71.5#185
15Gemma 4 31B69.8#382

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

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