SpecVQA - Chinese Descriptive (L0): leaderboard

SpecVQA (2026) asks questions about 620 scientific spectra (NMR, IR, XRD, Raman, mass, UV-Vis and XPS) from published papers, with expert-revised descriptive and reasoning questions in English and Chinese; GPT-o4-mini judges each answer against the reference, numeric answers within 5 percent. This board covers the Chinese descriptive (L0) questions. Accuracy in percent; higher is better.

Metric: Accuracy (0-1, times 100) on the Chinese descriptive (L0) questions; expert-revised question-answer pairs on 620 spectra (NMR, IR, XRD, Raman, MS, UV-Vis, XPS) from published papers; each answer judged correct or not by GPT-o4-mini against the reference, numeric answers within a 5 percent tolerance; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 Flash (Preview)80.47#152
2Gemini 3 Pro (Preview)79.39#119
3Gemini 2.5 Pro77.58#276
4Gemini 2.5 Flash74.73#463
5O4 Mini (2025-04-16)72.37#370
6GPT-571.54#178
7O3 (2025-04-16)70.27#231
8GPT-5.2Non-reasoning68.99#185
9GPT-5.1Non-reasoning68.99#240
10Claude Sonnet 4.5Thinking63.15#261
11Doubao-Seed-1.6-FlashThinking61.48#1250
12Claude Sonnet 4Thinking59.57#410
13Qwen 3 VL 8B Instruct59.27#1029
14Qwen 3 VL 8B (Thinking)58.05#910
15Doubao-Seed-1.6Thinking57.36

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

Interactive version: theaggregate.ai/benchmark?slug=specvqa-chinese-descriptive-l0 · How It Works · Data refreshed daily, snapshot 2026-11-21.