Same two news cases, Terra medium, three fresh original-input and three 0.15 compressed-input answers per case. Source-aware Terra-low and Sol-low validation.
| Case | Level | Decision | Context saved | Correct | Partial | Unresolved |
|---|---|---|---|---|---|---|
| Mali salaries | 0.30 | Protected fallback | 0.0% | 3/3 | 0 | 0 |
| Mali salaries | 0.25 | Protected fallback | 0.0% | 3/3 | 0 | 0 |
| Mali salaries | 0.15 | Protected fallback | 0.0% | 3/3 | 0 | 0 |
| VX quantifier | 0.30 | Accepted compression | 18.4% | 3/3 | 0 | 0 |
| VX quantifier | 0.25 | Accepted compression | 15.3% | 3/3 | 0 | 0 |
| VX quantifier | 0.15 | Accepted compression | 9.7% | 3/3 | 0 | 0 |
The integrity guard returned the original text at 0.15, preserving the explicit frozen-bank-account cause. All three answers and all three fresh controls were judged correct. This is successful protection with zero context savings; it is excluded from accepted-compression savings. The same protection occurred at 0.25 and 0.30.
At 0.15 the source retains one of two women in the key illness passage, clearer than the abbreviated one two women at 0.25 and 0.30. All three answers and all three fresh controls were judged correct. Context savings fell to 9.7%, compared with 15.3% at 0.25 and 18.4% at 0.30. There was no additional measured answer-quality benefit in these repeats.
Earlier levels use frozen follow-up results; 0.15 uses fresh matched controls. All counts are out of three repeats. Protected fallbacks are separate from accepted compressions. These two selected cases do not establish full-cohort quality or savings.