SpecVQA - Chinese Reasoning (L1): 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 reasoning (L1) questions. Accuracy in percent; higher is better.

Metric: Accuracy (0-1, times 100) on the Chinese reasoning (L1) 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)79#152
2Gemini 3 Pro (Preview)77.31#119
3Gemini 2.5 Pro75.71#276
4Gemini 2.5 Flash71.85#463
5GPT-5High71.47#178
6O3 (2025-04-16)70.43#231
7O4 Mini (2025-04-16)69.96#370
8GPT-5.1Non-reasoning62.9#240
9GPT-5.2Non-reasoning61.58#185
10Claude Sonnet 4.5Thinking57.82#261
11Doubao-Seed-1.6-FlashThinking55.74#1250
12Claude Sonnet 4Thinking55.65#410
13Qwen 3 VL 8B (Thinking)55.08#910
14Doubao-Seed-1.6Thinking54.52
15Qwen 3 Omni 30B A3B (Thinking)47.65

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-reasoning-l1 · How It Works · Data refreshed daily, snapshot 2026-11-21.