TeachObs - Scene Coding (Transcript): leaderboard
Classroom teaching-observation coding (TeachObs; Jeong et al., 2026): from the source-language transcript of a 15-second scene of a public K-12 lesson video, the model marks which of 39 observation codes (such as gesture, board work, questioning, feedback) apply. Macro F1 in percent over the codes against trained coders' gold labels, 1,099 scenes of six test lessons; higher is better.
Metric: Macro F1 (%; 39 observation codes per 15-second classroom scene from its transcript alone, 1,099 scenes of six test lessons). Source: arxiv.org. Saturation forecast: Around September 2028. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 (Low) | 27.7 | #75 (Claude Opus 4.7) |
| 2 | Gemini 3.1 Pro (Preview) (Low) | 26.2 | #91 (Gemini 3.1 Pro (Preview)) |
| 3 | GPT-5.5 (Low) | 25.4 | #48 (GPT-5.5) |
| 4 | Grok 4.3 (Low) | 25.1 | #332 (Grok 4.3) |
| 5 | GLM-5V Turbo | 23.6 | #342 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=teachobs-scene-coding-transcript · How It Works · Data refreshed daily, snapshot 2026-11-12.