HEx
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
HEx-PHI-FigStep
1Safety Score (1-ASR)
8
11x11 Hex
91.1Win rate
7
HEX n=11 (test)
3.943Side Length (L)
6
Hex 10k
96Best-Response Win Rate
5
HEX n=12 (test)
3.942Side Length (L)
5
Hex 15 × 15
71.88Winrate
4
Hex 13 × 13
76.12Winrate
4
Hex 9 × 9
70.5Winrate
4
Hex 7 × 7
60.25Winrate
4
Hex 5 × 5
49.25Winrate
4
HEX 24
5.485Side Length (L)
3
HEX 23
5.485Side Length (L)
3
HEX 22
5.286Side Length (L)
3
HEX 21
5Side Length (L)
3
HEX 20
5Side length (L)
3
HEX 19
4.619Side length (L)
3
HEX 18
4.619Side Length (L)
3
HEX 17
4.619Side Length (L)
3
HEX 16
4.528Side Length (L)
3
HEX 15
4.454Side length (L)
3
HEX 14
4.272Side Length (L)
3
HEX 13
4Side length (L)
3
HEX 30
6.005Side length (L)
2
HEX 29
6Side Length (L)
2
HEX 28
5.972Side length (L)
2