Flat-Pack Bench - Part Tracking: leaderboard
Metric: Track: recover the correspondence of shuffled part IDs between two segmented frames using the video, 257 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.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 32B (Thinking) | 45.53 |
| 2 | Qwen 2.5 VL 72B Instruct | 45.14 |
| 3 | InternVL3-78B | 42.02 |
| 4 | Qwen 3 VL 30B A3B Instruct | 42.02 |
| 5 | Qwen 3 VL 235B A22B Instruct | 39.69 |
| 6 | InternVL3-14B | 37.74 |
| 7 | Qwen 3 VL 4B (Thinking) | 37.74 |
| 8 | Qwen 3 VL 8B Instruct | 33.85 |
| 9 | Qwen 2.5 VL 7B Instruct | 33.07 |
| 10 | Qwen 2.5 VL 32B Instruct | 33.07 |
| 11 | Qwen 3 VL 4B Instruct | 32.68 |
| 12 | Qwen 3 VL 32B Instruct | 31.91 |
| 13 | GPT-5 | 25.68 |
| 14 | InternVL3-38B | 25.68 |
| 15 | Qwen 3 VL 8B (Thinking) | 25.29 |
Interactive version: theaggregate.ai/benchmark?slug=flat-pack-bench-part-tracking · How It Works · Data refreshed daily, snapshot 2026-10-07.