Long-term Video Object Segmentation v2
Models track object masks through long videos. The evaluation tests whether they can recover an object's identity after an extended absence.
- Made by
- Lingyi Hong and colleagues
- Released
- April 2024
- Kind
- Video understanding
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Prediction
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- New versions replace it
- Run by
- Anyone
- Based on
- A version of another benchmark
- In views
- Vision, Image, video and 3D generation
DeAOT-L tops the leaderboard with 65.9%
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
- 1 DeAOT-L Semi-supervised, LVOS training 65.9%
- 2 XMem Semi-supervised, LVOS training 65.2%
- 3 DeAOT-B Semi-supervised, LVOS training 62.8%
- 4 RDE Semi-supervised, LVOS training 62.1%
- 5 DEVA Unsupervised multiple objects, original weights 62%
- 6 LWL Semi-supervised, original weights 60.9%
- 7 STCN Semi-supervised, LVOS training 60.7%
- 8 CFBI Semi-supervised, LVOS training 58%
- 9 TMO Unsupervised single object, LVOS training 52.8%
- 10 FSNet Unsupervised single object, LVOS training 51%
- 11 Isomer Unsupervised single object, LVOS training 49.6%
- 12 HFAN Unsupervised single object, LVOS training 49.1%
- 13 AFB-URR Semi-supervised, LVOS training 48.7%
- 14 AMC-Net Unsupervised single object, original weights 44.6%
- 15 SimulFlow Unsupervised single object, original weights 44.1%
- 16 STEm-Seg Unsupervised multiple objects, LVOS training 39.6%
- 17 RVOS Unsupervised multiple objects, LVOS training 30.5%
- 18 SegGPT Semi-supervised, original weights 29.4%
- 19 SAM-PT Semi-supervised, original weights 22.1%
Source: The Long-term Video Object Segmentation v2 paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | DeAOT-L | 65.9% | Semi-supervised, LVOS training | Official leaderboard | |
| 2 | XMem | 65.2% | Semi-supervised, LVOS training | Official leaderboard | |
| 3 | DeAOT-B | 62.8% | Semi-supervised, LVOS training | Official leaderboard | |
| 4 | RDE | 62.1% | Semi-supervised, LVOS training | Official leaderboard | |
| 5 | DEVA | 62% | Unsupervised multiple objects, original weights | Official leaderboard | |
| 6 | LWL | 60.9% | Semi-supervised, original weights | Official leaderboard | |
| 7 | STCN | 60.7% | Semi-supervised, LVOS training | Official leaderboard | |
| 8 | CFBI | 58% | Semi-supervised, LVOS training | Official leaderboard | |
| 9 | TMO | 52.8% | Unsupervised single object, LVOS training | Official leaderboard | |
| 10 | FSNet | 51% | Unsupervised single object, LVOS training | Official leaderboard | |
| 11 | Isomer | 49.6% | Unsupervised single object, LVOS training | Official leaderboard | |
| 12 | HFAN | 49.1% | Unsupervised single object, LVOS training | Official leaderboard | |
| 13 | AFB-URR | 48.7% | Semi-supervised, LVOS training | Official leaderboard | |
| 14 | AMC-Net | 44.6% | Unsupervised single object, original weights | Official leaderboard | |
| 15 | SimulFlow | 44.1% | Unsupervised single object, original weights | Official leaderboard | |
| 16 | STEm-Seg | 39.6% | Unsupervised multiple objects, LVOS training | Official leaderboard | |
| 17 | RVOS | 30.5% | Unsupervised multiple objects, LVOS training | Official leaderboard | |
| 18 | SegGPT | 29.4% | Semi-supervised, original weights | Official leaderboard | |
| 19 | SAM-PT | 22.1% | Semi-supervised, original weights | Official leaderboard |
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy.
Qwen 3.8 Max 88.7%
- LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization.
Gemini 4 Argon 91.7%
- MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy.
Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy.
Qwen 3.8 Max 90.8%
- FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better.
Phi-4-multimodal 4.00%
- VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules.
Qwen3.5-Omni-Plus 93.1%
Data audit
All 8 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the benchmark’s paper (19 entries).
- Pass Models scored 19 entries on its leaderboard, none a model we track.
- Note Newest score No scores yet.
- Note Frontier models No models scored yet.
- Note Headroom No scores yet.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 19 entries at the last read, none a model we track.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.