Flat-Pack Bench - Part Mating: leaderboard
Flat-Pack Bench (2026) Part Mating questions (87) on furniture-assembly videos with segmented visual-prompt frames: decide whether two parts are connected in the final assembly. Multiple-choice accuracy in percent, each model scored in its best of six video and prompt settings; higher is better.
Metric: Mate: decide whether two parts are connected in the final assembly, 87 questions; multiple-choice accuracy (%, regex exact match) on furniture-assembly videos from IMaW with one or two segmented visual-prompt frames, zero-shot with greedy decoding (GPT-5 at its default); each row is the model's best of up to six settings (key-frame or trimmed video, mixed-media, collage or concat visual prompt) chosen by overall micro-average accuracy, the paper's main-table protocol; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 30 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 4B (Thinking) | 59.77 | #1354 |
| 2 | LLaVA-OneVision-72B | 57.47 | #1782 |
| 3 | Qwen 3 VL 4B Instruct | 56.32 | #1348 |
| 4 | LLaVA-Video-72B | 56.32 | #1211 |
| 5 | Qwen 2.5 VL 32B Instruct | 54.02 | #1065 |
| 6 | InternVL3-38B | 52.87 | #948 |
| 7 | LLaVA-Video-7B | 52.87 | #2303 |
| 8 | Video-LLaVA-7B | 51.72 | #3904 |
| 9 | Flat-Pack Bench VideoRefer (checkpoint unspecified) | 51.72 | |
| 10 | Gemini 3.1 Pro (Preview) | 49.43 | #102 |
| 11 | GPT-5 | 49.43 | #185 |
| 12 | Qwen 3 VL 30B A3B Instruct | 49.43 | #903 |
| 13 | InternVL3-14B | 48.28 | #1227 |
| 14 | Qwen 3 VL 32B (Thinking) | 47.13 | #577 |
| 15 | Qwen 3 VL 8B (Thinking) | 44.83 | #924 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=flat-pack-bench-part-mating · How It Works · Data refreshed daily, snapshot 2026-11-22.