Reconstruction Error Analysis on Agent-eval 8 templates user agent GPT-5.4 nano (aggregate)
5.6wMAPEM-Opt
Evaluation Results
| Method | Links | |
|---|---|---|
| M-Optε=0.1, B=3k + 12026.05 | 5.6 | |
| M-Optε=0.1, B=2k + 12026.05 | 7.6 | |
| M-Rootsε=0.1, B=3k + 12026.05 | 7.9 | |
| M-Optε=0.05, B=3k + 12026.05 | 10.5 | |
| M-Rootsε=0.1, B=2k + 12026.05 | 10.6 | |
| M-Optε=0.1, B=k + 12026.05 | 12.3 | |
| M-Rootsε=0.05, B=3k + 12026.05 | 13.9 | |
| M-Optε=0.05, B=2k + 12026.05 | 14 | |
| M-Rootsε=0.1, B=k + 12026.05 | 15.9 | |
| M-Rootsε=0.05, B=2k + 12026.05 | 17.8 | |
| M-Optε=0.05, B=k + 12026.05 | 20.9 | |
| M-Allε=0.1, B=k + 12026.05 | 21.6 | |
| M-Allε=0.1, B=3k + 12026.05 | 21.7 | |
| M-Allε=0.1, B=2k + 12026.05 | 22.2 | |
| M-Rootsε=0.05, B=k + 12026.05 | 26 | |
| M-Allε=0.05, B=k + 12026.05 | 32.8 | |
| M-Allε=0.05, B=2k + 12026.05 | 34.4 | |
| M-Allε=0.05, B=3k + 12026.05 | 34.6 | |
| M-Optε=0.01, B=3k + 12026.05 | 35.5 | |
| M-Rootsε=0.01, B=3k + 12026.05 | 43.2 | |
| M-Optε=0.01, B=2k + 12026.05 | 47.2 | |
| M-Rootsε=0.01, B=2k + 12026.05 | 55.5 | |
| M-Optε=0.01, B=k + 12026.05 | 71.6 | |
| M-Rootsε=0.01, B=k + 12026.05 | 82.6 | |
| M-Allε=0.01, B=k + 12026.05 | 98.8 | |
| M-Allε=0.01, B=2k + 12026.05 | 108 | |
| M-Allε=0.01, B=3k + 12026.05 | 115 |