OpenEQA
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
OpenEQA v1.0 (test)
67.7LLM-Match
32
OpenEQA v1 (ScanNet)
87.7LLM-Match
29
OpenEQA
69.3Accuracy
26
OpenEQA EM-EQA
86.8Accuracy
24
OpenEQA (test)
58.6Accuracy (< 30s)
18
OpenEQA
56.2LLM-Match
12
OpenEQA v1 (HM3D)
85.1LLM-Match
12
OpenEQA v1 (HM3D)
85.1LLM Match Score
12
OpenEQA EM-EQA Episodes up to 32 frames
86.8LLM-Match Score
10
OpenEQA
51.2Score
10
OpenEQA (test)
86.8GPT-4 Score
10
OpenEQA EM-EQA
86.8LLM-Match
8
OpenEQA 1/10 subset Unseen (val)
61.5LLM Match Score
7
OpenEQA v1 (test)
85.1Score
5
OpenEQA 184
52.6LLM-SR
4