WeRef-Bench
Models locate objects described in English or Chinese and return no boxes when the image contains no match.
Tencent reports mean F1 for matching objects and the rejection rate for absent objects. WeVisOne-VL-4B rejected 96.4% of absent targets.
- Made by
- Tencent
- Released
- October 2026
- Kind
- Detection and captioning
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Measures
- Capability
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision
- Strongest model tested
-
Qwen3.5-9B #128 on the capability index
- Last model added
- February 2026
- Scores refreshed
- November 21st, 2026
WeVisOne-VL-4B tops the leaderboard with 78.7% Qwen3.5-9B leads with 50.7%
Best score per entry from the benchmark's paper, read on November 21st, 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. The best model we track is Qwen3.5-9B, at 50.7%. This view shows only models we track, each at its best.
- 1 WeVisOne-VL-4B 78.7%
- 2 WeVisOne-VL-2B 73.2%
- 3 LocateAnything-3B 51.7%
- 4
Qwen3.5-9B 50.7%
- 5 Rex-Omni-3B 50.3%
- 6
Qwen3.5-4B 48.1%
- 7 Qwen3-VL-4B 46.3%
- 8 Qwen3-VL-8B 36.7%
- 9 Qwen3-VL-2B 30.8%
- 10 GroundingDINO-T 21.1%
Source: The WeRef-Bench paper
- 1
Qwen3.5-9B 50.7%
- 2
Qwen3.5-4B 48.1%
Source: The WeRef-Bench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | WeVisOne-VL-4B | 78.7% | Official leaderboard | ||
| 2 | WeVisOne-VL-2B | 73.2% | Official leaderboard | ||
| 3 | LocateAnything-3B | 51.7% | Official leaderboard | ||
| 4 | | 50.7% | Feb 2026 | Official leaderboard | |
| 5 | Rex-Omni-3B | 50.3% | Official leaderboard | ||
| 6 | | 48.1% | Feb 2026 | Official leaderboard | |
| 7 | Qwen3-VL-4B | 46.3% | Official leaderboard | ||
| 8 | Qwen3-VL-8B | 36.7% | Official leaderboard | ||
| 9 | Qwen3-VL-2B | 30.8% | Official leaderboard | ||
| 10 | GroundingDINO-T | 21.1% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 50.7% | Feb 2026 | Official leaderboard | |
| 2 | | 48.1% | Feb 2026 | Official leaderboard |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official WeRef-Bench 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 Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy.
Qwen 3.6 Plus 86%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy.
Seed2.1 Pro 90.7%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy.
Qwen 3.8 Max 97.7%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy.
Qwen 3.8 Max 85.9%
Data audit
8 of 10 checks pass, 2 warnings. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (10 entries).
- Pass Models scored 2 models we track, plus 8 other entries on its leaderboard.
- Warning Newest score The newest came 9 months ago, in February 2026, for Qwen3.5-4B.
- Note Frontier models None of the top 10 has a score. The strongest tested is Qwen3.5-9B (#128).
- Pass Headroom The best is 50.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 2 rows for models we track and 8 other entries at the last read.
- Pass Last read November 21st, 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.