PolyEdit multi-objective Hypervolume and MIP
Values are five-seed means with 95% t confidence intervals.
| Method | Hypervolume | MIP | Preference regret | Near-optimal | Unique |
|---|---|---|---|---|---|
| sft | 0.326 ± 0.009 | 0.554 ± 0.012 | 0.034 ± 0.012 | 0.782 ± 0.091 | 2.091 ± 0.176 |
| random | 0.328 ± 0.005 | 0.500 ± 0.000 | 0.088 ± 0.000 | 0.448 ± 0.019 | 4.409 ± 0.279 |
| greedy_verifier | 0.337 ± 0.004 | 0.568 ± 0.003 | 0.019 ± 0.003 | 0.833 ± 0.021 | 2.011 ± 0.129 |
| oracle_grid | 0.355 ± 0.000 | 0.588 ± 0.000 | 0.000 ± 0.000 | 1.000 ± 0.000 | 1.977 ± 0.000 |
Direction breakdown
| Direction | Method | Hypervolume | MIP |
|---|---|---|---|
| up_up | sft | 0.555 ± 0.006 | 0.717 ± 0.009 |
| up_up | random | 0.530 ± 0.011 | 0.623 ± 0.004 |
| up_up | greedy_verifier | 0.542 ± 0.010 | 0.710 ± 0.005 |
| up_up | oracle_grid | 0.572 ± 0.000 | 0.731 ± 0.000 |
| up_down | sft | 0.266 ± 0.010 | 0.532 ± 0.005 |
| up_down | random | 0.277 ± 0.004 | 0.499 ± 0.004 |
| up_down | greedy_verifier | 0.281 ± 0.004 | 0.549 ± 0.002 |
| up_down | oracle_grid | 0.300 ± 0.000 | 0.562 ± 0.000 |
| down_up | sft | 0.268 ± 0.017 | 0.534 ± 0.017 |
| down_up | random | 0.274 ± 0.004 | 0.501 ± 0.004 |
| down_up | greedy_verifier | 0.283 ± 0.009 | 0.549 ± 0.004 |
| down_up | oracle_grid | 0.286 ± 0.000 | 0.569 ± 0.000 |
| down_down | sft | 0.216 ± 0.015 | 0.433 ± 0.020 |
| down_down | random | 0.231 ± 0.007 | 0.377 ± 0.004 |
| down_down | greedy_verifier | 0.244 ± 0.003 | 0.466 ± 0.003 |
| down_down | oracle_grid | 0.261 ± 0.000 | 0.489 ± 0.000 |