All 18 author-supplied cases are included. Each case contains all available text logs and uniformly selected original PNG frames. Click a case to inspect the observations, value maps, navigation maps, and prompts/responses.
These selected historical cases are qualitative material. Dataset split, model version, benchmark success and SPL are not established by these text files. They are not paired comparisons or evidence from the repaired implementation. Error-marked directories and missing text records are preserved in the index.
| Case | Target | Step directories | Text logs | Error-marked |
|---|---|---|---|---|
| 115_853 | sofa | 6 | 6 | 0 |
| 160_853 | tv screen | 7 | 7 | 1 |
| 199_890 | plant | 29 | 29 | 2 |
| 220_890 | plant | 40 | 40 | 1 |
| 249_890 | tv screen | 40 | 40 | 7 |
| 265_890 | toilet | 19 | 19 | 2 |
| 289_890 | chair | 40 | 40 | 5 |
| 297_843 | toilet | 8 | 8 | 0 |
| 2_877 | bed | 4 | 4 | 0 |
| 310_843 | sofa | 2 | 2 | 0 |
| 32_877 | chair | 6 | 6 | 0 |
| 352_843 | toilet | 15 | 15 | 1 |
| 382_843 | bed | 24 | 24 | 0 |
| 399_824 | bed | 14 | 14 | 0 |
| 450_824 | bed | 40 | 40 | 2 |
| 546_880 | chair | 4 | 4 | 0 |
| 594_829 | chair | 10 | 9 | 0 |
| 63_877 | sofa | 30 | 30 | 1 |