Flat-Pack Bench - Part Tracking: leaderboard
Flat-Pack Bench (2026) Part Tracking questions (257) on furniture-assembly videos with segmented visual-prompt frames: recover the correspondence of shuffled part IDs between two segmented frames using the video. Multiple-choice accuracy in percent, each model scored in its best of six video and prompt settings; higher is better.
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. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 32B (Thinking) | 45.53 | #573 |
| 2 | Qwen 2.5 VL 72B Instruct | 45.14 | #841 |
| 3 | Perception-LM-8B | 44.75 | #932 |
| 4 | InternVL3-78B | 42.02 | #759 |
| 5 | Qwen 3 VL 30B A3B Instruct | 42.02 | #895 |
| 6 | Qwen 3 VL 235B A22B Instruct | 39.69 | #532 |
| 7 | LLaVA-OneVision-72B | 38.91 | #1771 |
| 8 | InternVL3-14B | 37.74 | #1214 |
| 9 | Qwen 3 VL 4B (Thinking) | 37.74 | #1334 |
| 10 | LLaVA-Video-72B | 35.8 | #1193 |
| 11 | LLaVA-OneVision-7B | 34.24 | #2981 |
| 12 | Qwen 3 VL 8B Instruct | 33.85 | #1029 |
| 13 | Qwen 2.5 VL 7B Instruct | 33.07 | #1902 |
| 14 | Qwen 2.5 VL 32B Instruct | 33.07 | #1062 |
| 15 | Qwen 3 VL 4B Instruct | 32.68 | #1332 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=flat-pack-bench-part-tracking · How It Works · Data refreshed daily, snapshot 2026-11-21.