SYNCR - Object Re-identification: leaderboard

Accuracy (%) on matching an object seen in different videos in SYNCR (2026), a synthetic benchmark of multi-video questions with programmatically verified answers; zero-shot four-option multiple choice, exact match, 200 questions per task (chance 25). Higher is better.

Metric: Exact-match accuracy (%) on the SYNCR Object Re-identification task (matching an object seen in different videos), zero-shot four-option multiple choice over 2 to 4 labelled input videos, exact match, deterministic decoding, on the paper's balanced subset of 200 questions per task; chance is 25; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash64#175
2LLaVA-OV-72B50#1879
3Qwen 3 VL 32B Instruct44.5#606
4Qwen 2.5 VL 32B Instruct41.5#1062
5InternVL3.5-14B38#1504
6LLaVA-OV-7B37.5#4475
7GPT-5.436#135
8Qwen 3 VL 8B Instruct33#1029
9InternVL3.5-4B32.5#1930
10Qwen 3 VL 2B Instruct32#2367
11InternVL3.5-2B31#3284
12LLaVA-OV-0.5B31#4352
13InternVL3.5-1B30.5#6112
14Qwen 2.5 VL 72B Instruct30#841
15InternVL3.5-38B29#1021

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=syncr-object-re-identification · How It Works · Data refreshed daily, snapshot 2026-11-21.