Benchmark Almanac

SPair-71k

CATs 49.9% tops its leaderboard
Not in model cards

Models match corresponding parts of objects across image pairs. The benchmark includes changes in viewpoint, scale, and occlusion.

Made by
Juhong Min and colleagues
Released
August 2019
Kind
Spatial reasoning and counting
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Images → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
In views
Vision
18
Entries on SOTA2’s leaderboard, 0 models we track

CATs tops the leaderboard with 49.9%

Best score per entry from SOTA2’s leaderboard, read on November 20th, 2026. 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. 1 CATs Feat.-level=Multi, FT. feat.=true, Aggregation=Transformer 49.9%
  2. 2 CHM Feat.-level=Single, FT. feat.=false, Aggregation=6D Conv. 46.3%
  3. 3 DHPF Feat.-level=Multi, FT. feat.=false, Aggregation=RHM 37.3%
  4. 4 GSF Feat.-level=Multi, FT. feat.=false, Aggregation=2D Conv. 36.1%
  5. 5 SCOT Feat.-level=Multi, Aggregation=OT-RHM 35.6%
  6. 6 ANC-Net Backbone=ResNeXt-101 30.1%
  7. 7 HPF Backbone=ResNet-101 28.2%
  8. 8 DCCNet Backbone=ResNet-101 26.7%
  9. 9 NC-Net Backbone=ResNet-101 26.4%
  10. 10 SFNet Backbone=ResNet-101 26%
  11. 11 RTNs Feat.-level=Single, FT. feat.=false, Aggregation=- 25.7%
  12. 12 WTA Feat.-level=Single, FT. feat.=false, Aggregation=- 25.7%
  13. 13 A2Net Feat.-level=Single, FT. feat.=false, Aggregation=- 22.3%
  14. 14 Weakalign Backbone=ResNet-101 21.1%
  15. 15 WeakAlign Feat.-level=Single, FT. feat.=false, Aggregation=- 20.9%
  16. 16 CNNGeo Feat.-level=Single, FT. feat.=false, Aggregation=- 20.6%
  17. 17 UCN Backbone=ResNet-101 17.7%
  18. 18 Identity mapping 3.70%

Source: SOTA2’s leaderboard

#EntryScoreSettingDateSource
1 CATs 49.9% Feat.-level=Multi, FT. feat.=true, Aggregation=Transformer Official leaderboard
2 CHM 46.3% Feat.-level=Single, FT. feat.=false, Aggregation=6D Conv. Official leaderboard
3 DHPF 37.3% Feat.-level=Multi, FT. feat.=false, Aggregation=RHM Official leaderboard
4 GSF 36.1% Feat.-level=Multi, FT. feat.=false, Aggregation=2D Conv. Official leaderboard
5 SCOT 35.6% Feat.-level=Multi, Aggregation=OT-RHM Official leaderboard
6 ANC-Net 30.1% Backbone=ResNeXt-101 Official leaderboard
7 HPF 28.2% Backbone=ResNet-101 Official leaderboard
8 DCCNet 26.7% Backbone=ResNet-101 Official leaderboard
9 NC-Net 26.4% Backbone=ResNet-101 Official leaderboard
10 SFNet 26% Backbone=ResNet-101 Official leaderboard
11 RTNs 25.7% Feat.-level=Single, FT. feat.=false, Aggregation=- Official leaderboard
12 WTA 25.7% Feat.-level=Single, FT. feat.=false, Aggregation=- Official leaderboard
13 A2Net 22.3% Feat.-level=Single, FT. feat.=false, Aggregation=- Official leaderboard
14 Weakalign 21.1% Backbone=ResNet-101 Official leaderboard
15 WeakAlign 20.9% Feat.-level=Single, FT. feat.=false, Aggregation=- Official leaderboard
16 CNNGeo 20.6% Feat.-level=Single, FT. feat.=false, Aggregation=- Official leaderboard
17 UCN 17.7% Backbone=ResNet-101 Official leaderboard
18 Identity mapping 3.70% 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 Vision

All 636 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

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 official leaderboard (18 entries).
  • Pass Models scored 18 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 SOTA2: 18 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.

How we check