Add final 64-sample MBPP evaluation
Browse files- eval64/REPORT.md +42 -0
- eval64/comparisons_compact.json +987 -0
- eval64/manifest.json +107 -0
- eval64/metrics_compact.json +766 -0
- eval64/summary.csv +5 -0
eval64/REPORT.md
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Five-round self-evolution: final 64-sample evaluation
|
| 2 |
+
|
| 3 |
+
This is the supplemental final-checkpoint evaluation for the single-seed five-round run. It evaluates the base model and round-5 `plain`, `spd_hard`, and `spectral_soft` checkpoints on all 500 held-out MBPP tasks with 64 common-policy samples per task (32,000 samples per model).
|
| 4 |
+
|
| 5 |
+
## Results
|
| 6 |
+
|
| 7 |
+
| Model | pass@1 | pass@4 | pass@8 | pass@16 | pass@64 | total coverage@4 | total coverage@16 | total coverage@64 |
|
| 8 |
+
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 9 |
+
| base | 0.379 | 0.551 | 0.611 | 0.657 | 0.720 | 1.267 | 4.040 | 12.802 |
|
| 10 |
+
| plain | 0.414 | 0.565 | 0.619 | 0.656 | 0.704 | 1.166 | 3.159 | 8.506 |
|
| 11 |
+
| spd_hard | 0.418 | 0.566 | 0.619 | 0.656 | 0.702 | 1.153 | 3.103 | 8.388 |
|
| 12 |
+
| spectral_soft | 0.403 | 0.562 | 0.619 | 0.662 | 0.726 | 1.244 | 3.758 | 11.510 |
|
| 13 |
+
|
| 14 |
+
Correct-matched implementation coverage conditions on drawing the stated number of correct samples:
|
| 15 |
+
|
| 16 |
+
| Model | correct coverage@4 (eligible) | @8 (eligible) | @16 (eligible) | @64 (eligible) |
|
| 17 |
+
| --- | ---: | ---: | ---: | ---: |
|
| 18 |
+
| base | 3.358 (322/500) | 5.987 (294/500) | 10.423 (251/500) | 19.875 (24/500) |
|
| 19 |
+
| plain | 2.856 (321/500) | 4.690 (305/500) | 7.324 (263/500) | 8.686 (51/500) |
|
| 20 |
+
| spd_hard | 2.810 (321/500) | 4.590 (303/500) | 7.136 (264/500) | 7.981 (53/500) |
|
| 21 |
+
| spectral_soft | 3.143 (323/500) | 5.451 (302/500) | 9.091 (259/500) | 12.176 (34/500) |
|
| 22 |
+
|
| 23 |
+
## Main findings
|
| 24 |
+
|
| 25 |
+
- Relative to the intended hard-projection control (`spd_hard`), `spectral_soft` restores implementation coverage: total coverage improves by +0.091 [+0.066, +0.116] at draw budget 4 and +3.122 [+2.686, +3.596] at budget 64. Correct-matched coverage improves by +0.334 [+0.290, +0.379] at four correct draws.
|
| 26 |
+
- `spectral_soft` also improves pass@64 over `spd_hard` by +0.024 [+0.006, +0.044]. Its pass@1 is lower by -0.014 [-0.019, -0.010]. The correctness delta fails the preregistered absolute 0.01 noninferiority margin (`noninferior=false`).
|
| 27 |
+
- Relative to `plain`, `spectral_soft` gains +3.004 [+2.568, +3.494] total coverage@64 and +0.295 [+0.249, +0.343] correct-matched coverage@4, with pass@1 -0.011 [-0.016, -0.006].
|
| 28 |
+
- Relative to the base model, `spectral_soft` improves pass@1 by +0.024 [+0.016, +0.032] and has an inconclusive pass@64 difference of +0.006 [-0.018, +0.030]. It does not fully retain base diversity: total coverage@64 differs by -1.292 [-1.786, -0.780].
|
| 29 |
+
|
| 30 |
+
## Interpretation and limits
|
| 31 |
+
|
| 32 |
+
- The result supports the mechanism claim that soft spectral attenuation preserves substantially more implementation diversity than the hard projection after five rounds. It does not meet the stated correctness noninferiority gate against `plain` or `spd_hard` at the 0.01 margin.
|
| 33 |
+
- `spd_hard` is an aggressive hard-projection control; its lower coverage is expected and is evidence about the cost of hard removal, rather than an optimization failure.
|
| 34 |
+
- Correct-matched coverage@64 has small and model-dependent eligibility (24–53 tasks; only 30 paired tasks for `spectral_soft` versus `spd_hard`). Treat it as a sparse conditional diagnostic. Budgets 4, 8, and 16 are more stable.
|
| 35 |
+
- This run uses one training seed (43). Bootstrap intervals resample the 500 evaluation tasks and do not measure training-seed variation.
|
| 36 |
+
- AST fingerprints are implementation proxies, not audited semantic algorithm labels. Local execution used the recorded Python 3.11 task-test harness; all four models have identical recorded evaluator and task-harness provenance.
|
| 37 |
+
|
| 38 |
+
## Reproducibility
|
| 39 |
+
|
| 40 |
+
- Training: 291 MBPP training tasks × 16 raw completions per task × 5 self-evolution rounds; one LoRA epoch per round.
|
| 41 |
+
- Evaluation: seed 43, temperature 0.8, top-p 0.95, 64 samples per each of 500 held-out tasks.
|
| 42 |
+
- Exact aggregate values and task-bootstrap intervals are in `summary.csv`; compact paired results are in `comparisons_compact.json`; source-file hashes are in `manifest.json`.
|
eval64/comparisons_compact.json
ADDED
|
@@ -0,0 +1,987 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"plain_minus_base": {
|
| 3 |
+
"direction": "current_minus_previous",
|
| 4 |
+
"task_count": 500,
|
| 5 |
+
"evaluation_identity_check": "declared_provenance_equal",
|
| 6 |
+
"correctness": {
|
| 7 |
+
"metric": "task_macro_correct_fraction",
|
| 8 |
+
"margin": 0.01,
|
| 9 |
+
"delta": {
|
| 10 |
+
"mean": 0.03503125,
|
| 11 |
+
"all_task_mean": 0.03503125,
|
| 12 |
+
"ci95": [
|
| 13 |
+
0.025403906249999997,
|
| 14 |
+
0.04540625
|
| 15 |
+
],
|
| 16 |
+
"eligible_tasks": 500,
|
| 17 |
+
"total_tasks": 500,
|
| 18 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 19 |
+
"bootstrap_valid_replicates": 2000
|
| 20 |
+
},
|
| 21 |
+
"noninferior": true
|
| 22 |
+
},
|
| 23 |
+
"pass_at_k": {
|
| 24 |
+
"1": {
|
| 25 |
+
"mean": 0.03503125,
|
| 26 |
+
"all_task_mean": 0.03503125,
|
| 27 |
+
"ci95": [
|
| 28 |
+
0.025403906249999997,
|
| 29 |
+
0.04540625
|
| 30 |
+
],
|
| 31 |
+
"eligible_tasks": 500,
|
| 32 |
+
"total_tasks": 500,
|
| 33 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 34 |
+
"bootstrap_valid_replicates": 2000
|
| 35 |
+
},
|
| 36 |
+
"4": {
|
| 37 |
+
"mean": 0.014645705849764554,
|
| 38 |
+
"all_task_mean": 0.014645705849764554,
|
| 39 |
+
"ci95": [
|
| 40 |
+
0.0035726654768200275,
|
| 41 |
+
0.02578586451172219
|
| 42 |
+
],
|
| 43 |
+
"eligible_tasks": 500,
|
| 44 |
+
"total_tasks": 500,
|
| 45 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 46 |
+
"bootstrap_valid_replicates": 2000
|
| 47 |
+
},
|
| 48 |
+
"8": {
|
| 49 |
+
"mean": 0.007597980982169215,
|
| 50 |
+
"all_task_mean": 0.007597980982169215,
|
| 51 |
+
"ci95": [
|
| 52 |
+
-0.004422959510671402,
|
| 53 |
+
0.019885941463066457
|
| 54 |
+
],
|
| 55 |
+
"eligible_tasks": 500,
|
| 56 |
+
"total_tasks": 500,
|
| 57 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 58 |
+
"bootstrap_valid_replicates": 2000
|
| 59 |
+
},
|
| 60 |
+
"16": {
|
| 61 |
+
"mean": -0.000560707010670496,
|
| 62 |
+
"all_task_mean": -0.000560707010670496,
|
| 63 |
+
"ci95": [
|
| 64 |
+
-0.013798564385109827,
|
| 65 |
+
0.012829319603621431
|
| 66 |
+
],
|
| 67 |
+
"eligible_tasks": 500,
|
| 68 |
+
"total_tasks": 500,
|
| 69 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 70 |
+
"bootstrap_valid_replicates": 2000
|
| 71 |
+
},
|
| 72 |
+
"64": {
|
| 73 |
+
"mean": -0.016,
|
| 74 |
+
"all_task_mean": -0.016,
|
| 75 |
+
"ci95": [
|
| 76 |
+
-0.04,
|
| 77 |
+
0.008
|
| 78 |
+
],
|
| 79 |
+
"eligible_tasks": 500,
|
| 80 |
+
"total_tasks": 500,
|
| 81 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 82 |
+
"bootstrap_valid_replicates": 2000
|
| 83 |
+
}
|
| 84 |
+
},
|
| 85 |
+
"implementation_coverage_at_k": {
|
| 86 |
+
"1": {
|
| 87 |
+
"mean": 0.03503125000000001,
|
| 88 |
+
"all_task_mean": 0.03503125000000001,
|
| 89 |
+
"ci95": [
|
| 90 |
+
0.025403906250000007,
|
| 91 |
+
0.04540625
|
| 92 |
+
],
|
| 93 |
+
"eligible_tasks": 500,
|
| 94 |
+
"total_tasks": 500,
|
| 95 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 96 |
+
"bootstrap_valid_replicates": 2000
|
| 97 |
+
},
|
| 98 |
+
"4": {
|
| 99 |
+
"mean": -0.10042582030167964,
|
| 100 |
+
"all_task_mean": -0.10042582030167964,
|
| 101 |
+
"ci95": [
|
| 102 |
+
-0.14195690591083074,
|
| 103 |
+
-0.058621159282062935
|
| 104 |
+
],
|
| 105 |
+
"eligible_tasks": 500,
|
| 106 |
+
"total_tasks": 500,
|
| 107 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 108 |
+
"bootstrap_valid_replicates": 2000
|
| 109 |
+
},
|
| 110 |
+
"8": {
|
| 111 |
+
"mean": -0.34413941183365204,
|
| 112 |
+
"all_task_mean": -0.34413941183365204,
|
| 113 |
+
"ci95": [
|
| 114 |
+
-0.42891005746967154,
|
| 115 |
+
-0.2601543456124548
|
| 116 |
+
],
|
| 117 |
+
"eligible_tasks": 500,
|
| 118 |
+
"total_tasks": 500,
|
| 119 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 120 |
+
"bootstrap_valid_replicates": 2000
|
| 121 |
+
},
|
| 122 |
+
"16": {
|
| 123 |
+
"mean": -0.8804222992412762,
|
| 124 |
+
"all_task_mean": -0.8804222992412762,
|
| 125 |
+
"ci95": [
|
| 126 |
+
-1.050281263759979,
|
| 127 |
+
-0.715082851722488
|
| 128 |
+
],
|
| 129 |
+
"eligible_tasks": 500,
|
| 130 |
+
"total_tasks": 500,
|
| 131 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 132 |
+
"bootstrap_valid_replicates": 2000
|
| 133 |
+
},
|
| 134 |
+
"64": {
|
| 135 |
+
"mean": -4.296,
|
| 136 |
+
"all_task_mean": -4.296,
|
| 137 |
+
"ci95": [
|
| 138 |
+
-4.91805,
|
| 139 |
+
-3.6919500000000003
|
| 140 |
+
],
|
| 141 |
+
"eligible_tasks": 500,
|
| 142 |
+
"total_tasks": 500,
|
| 143 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 144 |
+
"bootstrap_valid_replicates": 2000
|
| 145 |
+
}
|
| 146 |
+
},
|
| 147 |
+
"correct_matched_coverage_at_budgets": {
|
| 148 |
+
"4": {
|
| 149 |
+
"mean": -0.5105450460781567,
|
| 150 |
+
"all_task_mean": null,
|
| 151 |
+
"ci95": [
|
| 152 |
+
-0.5691189932070015,
|
| 153 |
+
-0.45464704799623973
|
| 154 |
+
],
|
| 155 |
+
"eligible_tasks": 311,
|
| 156 |
+
"total_tasks": 500,
|
| 157 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 158 |
+
"bootstrap_valid_replicates": 2000
|
| 159 |
+
},
|
| 160 |
+
"8": {
|
| 161 |
+
"mean": -1.3638615453142144,
|
| 162 |
+
"all_task_mean": null,
|
| 163 |
+
"ci95": [
|
| 164 |
+
-1.4887791049135546,
|
| 165 |
+
-1.2436493522093226
|
| 166 |
+
],
|
| 167 |
+
"eligible_tasks": 285,
|
| 168 |
+
"total_tasks": 500,
|
| 169 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 170 |
+
"bootstrap_valid_replicates": 2000
|
| 171 |
+
},
|
| 172 |
+
"16": {
|
| 173 |
+
"mean": -3.1732880253548577,
|
| 174 |
+
"all_task_mean": null,
|
| 175 |
+
"ci95": [
|
| 176 |
+
-3.4222336608574406,
|
| 177 |
+
-2.934467199126611
|
| 178 |
+
],
|
| 179 |
+
"eligible_tasks": 237,
|
| 180 |
+
"total_tasks": 500,
|
| 181 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 182 |
+
"bootstrap_valid_replicates": 2000
|
| 183 |
+
},
|
| 184 |
+
"64": {
|
| 185 |
+
"mean": -11.318181818181818,
|
| 186 |
+
"all_task_mean": null,
|
| 187 |
+
"ci95": [
|
| 188 |
+
-14.392980072463768,
|
| 189 |
+
-8.49953703703704
|
| 190 |
+
],
|
| 191 |
+
"eligible_tasks": 22,
|
| 192 |
+
"total_tasks": 500,
|
| 193 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 194 |
+
"bootstrap_valid_replicates": 2000
|
| 195 |
+
}
|
| 196 |
+
},
|
| 197 |
+
"caveat": "Metric, sampling budgets and available declared evaluator/task harness identities were checked. Missing legacy provenance is explicitly unverified. Identical prompts, decoding protocol, and stable strategy annotation identities remain caller responsibilities. Partial-eligibility differences describe only the reported paired subset; these pointwise CIs do not establish an overall diversity gain."
|
| 198 |
+
},
|
| 199 |
+
"spd_hard_minus_base": {
|
| 200 |
+
"direction": "current_minus_previous",
|
| 201 |
+
"task_count": 500,
|
| 202 |
+
"evaluation_identity_check": "declared_provenance_equal",
|
| 203 |
+
"correctness": {
|
| 204 |
+
"metric": "task_macro_correct_fraction",
|
| 205 |
+
"margin": 0.01,
|
| 206 |
+
"delta": {
|
| 207 |
+
"mean": 0.03821875,
|
| 208 |
+
"all_task_mean": 0.03821875,
|
| 209 |
+
"ci95": [
|
| 210 |
+
0.028715625,
|
| 211 |
+
0.04859375
|
| 212 |
+
],
|
| 213 |
+
"eligible_tasks": 500,
|
| 214 |
+
"total_tasks": 500,
|
| 215 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 216 |
+
"bootstrap_valid_replicates": 2000
|
| 217 |
+
},
|
| 218 |
+
"noninferior": true
|
| 219 |
+
},
|
| 220 |
+
"pass_at_k": {
|
| 221 |
+
"1": {
|
| 222 |
+
"mean": 0.03821875,
|
| 223 |
+
"all_task_mean": 0.03821875,
|
| 224 |
+
"ci95": [
|
| 225 |
+
0.028715625,
|
| 226 |
+
0.04859375
|
| 227 |
+
],
|
| 228 |
+
"eligible_tasks": 500,
|
| 229 |
+
"total_tasks": 500,
|
| 230 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 231 |
+
"bootstrap_valid_replicates": 2000
|
| 232 |
+
},
|
| 233 |
+
"4": {
|
| 234 |
+
"mean": 0.01520221097428924,
|
| 235 |
+
"all_task_mean": 0.01520221097428924,
|
| 236 |
+
"ci95": [
|
| 237 |
+
0.0037094430699302454,
|
| 238 |
+
0.02625112453098637
|
| 239 |
+
],
|
| 240 |
+
"eligible_tasks": 500,
|
| 241 |
+
"total_tasks": 500,
|
| 242 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 243 |
+
"bootstrap_valid_replicates": 2000
|
| 244 |
+
},
|
| 245 |
+
"8": {
|
| 246 |
+
"mean": 0.007794947373959033,
|
| 247 |
+
"all_task_mean": 0.007794947373959033,
|
| 248 |
+
"ci95": [
|
| 249 |
+
-0.004214445501553613,
|
| 250 |
+
0.019853462255389462
|
| 251 |
+
],
|
| 252 |
+
"eligible_tasks": 500,
|
| 253 |
+
"total_tasks": 500,
|
| 254 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 255 |
+
"bootstrap_valid_replicates": 2000
|
| 256 |
+
},
|
| 257 |
+
"16": {
|
| 258 |
+
"mean": -0.00046338000834429194,
|
| 259 |
+
"all_task_mean": -0.00046338000834429194,
|
| 260 |
+
"ci95": [
|
| 261 |
+
-0.014188135926675798,
|
| 262 |
+
0.01320068172292071
|
| 263 |
+
],
|
| 264 |
+
"eligible_tasks": 500,
|
| 265 |
+
"total_tasks": 500,
|
| 266 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 267 |
+
"bootstrap_valid_replicates": 2000
|
| 268 |
+
},
|
| 269 |
+
"64": {
|
| 270 |
+
"mean": -0.018,
|
| 271 |
+
"all_task_mean": -0.018,
|
| 272 |
+
"ci95": [
|
| 273 |
+
-0.042,
|
| 274 |
+
0.004
|
| 275 |
+
],
|
| 276 |
+
"eligible_tasks": 500,
|
| 277 |
+
"total_tasks": 500,
|
| 278 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 279 |
+
"bootstrap_valid_replicates": 2000
|
| 280 |
+
}
|
| 281 |
+
},
|
| 282 |
+
"implementation_coverage_at_k": {
|
| 283 |
+
"1": {
|
| 284 |
+
"mean": 0.03821875000000001,
|
| 285 |
+
"all_task_mean": 0.03821875000000001,
|
| 286 |
+
"ci95": [
|
| 287 |
+
0.02871562500000001,
|
| 288 |
+
0.04859375
|
| 289 |
+
],
|
| 290 |
+
"eligible_tasks": 500,
|
| 291 |
+
"total_tasks": 500,
|
| 292 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 293 |
+
"bootstrap_valid_replicates": 2000
|
| 294 |
+
},
|
| 295 |
+
"4": {
|
| 296 |
+
"mean": -0.11387302636549065,
|
| 297 |
+
"all_task_mean": -0.11387302636549065,
|
| 298 |
+
"ci95": [
|
| 299 |
+
-0.15733176276724334,
|
| 300 |
+
-0.07090118126904384
|
| 301 |
+
],
|
| 302 |
+
"eligible_tasks": 500,
|
| 303 |
+
"total_tasks": 500,
|
| 304 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 305 |
+
"bootstrap_valid_replicates": 2000
|
| 306 |
+
},
|
| 307 |
+
"8": {
|
| 308 |
+
"mean": -0.37528642309145643,
|
| 309 |
+
"all_task_mean": -0.37528642309145643,
|
| 310 |
+
"ci95": [
|
| 311 |
+
-0.4643529187722048,
|
| 312 |
+
-0.28801556663024797
|
| 313 |
+
],
|
| 314 |
+
"eligible_tasks": 500,
|
| 315 |
+
"total_tasks": 500,
|
| 316 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 317 |
+
"bootstrap_valid_replicates": 2000
|
| 318 |
+
},
|
| 319 |
+
"16": {
|
| 320 |
+
"mean": -0.9362489391249798,
|
| 321 |
+
"all_task_mean": -0.9362489391249798,
|
| 322 |
+
"ci95": [
|
| 323 |
+
-1.1056935955785272,
|
| 324 |
+
-0.7672233437329382
|
| 325 |
+
],
|
| 326 |
+
"eligible_tasks": 500,
|
| 327 |
+
"total_tasks": 500,
|
| 328 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 329 |
+
"bootstrap_valid_replicates": 2000
|
| 330 |
+
},
|
| 331 |
+
"64": {
|
| 332 |
+
"mean": -4.414,
|
| 333 |
+
"all_task_mean": -4.414,
|
| 334 |
+
"ci95": [
|
| 335 |
+
-5.0520499999999995,
|
| 336 |
+
-3.798
|
| 337 |
+
],
|
| 338 |
+
"eligible_tasks": 500,
|
| 339 |
+
"total_tasks": 500,
|
| 340 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 341 |
+
"bootstrap_valid_replicates": 2000
|
| 342 |
+
}
|
| 343 |
+
},
|
| 344 |
+
"correct_matched_coverage_at_budgets": {
|
| 345 |
+
"4": {
|
| 346 |
+
"mean": -0.5602393104834131,
|
| 347 |
+
"all_task_mean": null,
|
| 348 |
+
"ci95": [
|
| 349 |
+
-0.620203311960209,
|
| 350 |
+
-0.5024146068264466
|
| 351 |
+
],
|
| 352 |
+
"eligible_tasks": 311,
|
| 353 |
+
"total_tasks": 500,
|
| 354 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 355 |
+
"bootstrap_valid_replicates": 2000
|
| 356 |
+
},
|
| 357 |
+
"8": {
|
| 358 |
+
"mean": -1.4580467687163405,
|
| 359 |
+
"all_task_mean": null,
|
| 360 |
+
"ci95": [
|
| 361 |
+
-1.5912065108858593,
|
| 362 |
+
-1.328951991999187
|
| 363 |
+
],
|
| 364 |
+
"eligible_tasks": 285,
|
| 365 |
+
"total_tasks": 500,
|
| 366 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 367 |
+
"bootstrap_valid_replicates": 2000
|
| 368 |
+
},
|
| 369 |
+
"16": {
|
| 370 |
+
"mean": -3.3915337280063698,
|
| 371 |
+
"all_task_mean": null,
|
| 372 |
+
"ci95": [
|
| 373 |
+
-3.6596124817747016,
|
| 374 |
+
-3.1355820916237627
|
| 375 |
+
],
|
| 376 |
+
"eligible_tasks": 237,
|
| 377 |
+
"total_tasks": 500,
|
| 378 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 379 |
+
"bootstrap_valid_replicates": 2000
|
| 380 |
+
},
|
| 381 |
+
"64": {
|
| 382 |
+
"mean": -12.333333333333334,
|
| 383 |
+
"all_task_mean": null,
|
| 384 |
+
"ci95": [
|
| 385 |
+
-15.476190476190476,
|
| 386 |
+
-9.449861111111112
|
| 387 |
+
],
|
| 388 |
+
"eligible_tasks": 21,
|
| 389 |
+
"total_tasks": 500,
|
| 390 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 391 |
+
"bootstrap_valid_replicates": 2000
|
| 392 |
+
}
|
| 393 |
+
},
|
| 394 |
+
"caveat": "Metric, sampling budgets and available declared evaluator/task harness identities were checked. Missing legacy provenance is explicitly unverified. Identical prompts, decoding protocol, and stable strategy annotation identities remain caller responsibilities. Partial-eligibility differences describe only the reported paired subset; these pointwise CIs do not establish an overall diversity gain."
|
| 395 |
+
},
|
| 396 |
+
"spectral_soft_minus_base": {
|
| 397 |
+
"direction": "current_minus_previous",
|
| 398 |
+
"task_count": 500,
|
| 399 |
+
"evaluation_identity_check": "declared_provenance_equal",
|
| 400 |
+
"correctness": {
|
| 401 |
+
"metric": "task_macro_correct_fraction",
|
| 402 |
+
"margin": 0.01,
|
| 403 |
+
"delta": {
|
| 404 |
+
"mean": 0.023875,
|
| 405 |
+
"all_task_mean": 0.023875,
|
| 406 |
+
"ci95": [
|
| 407 |
+
0.01578046875,
|
| 408 |
+
0.03222187499999998
|
| 409 |
+
],
|
| 410 |
+
"eligible_tasks": 500,
|
| 411 |
+
"total_tasks": 500,
|
| 412 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 413 |
+
"bootstrap_valid_replicates": 2000
|
| 414 |
+
},
|
| 415 |
+
"noninferior": true
|
| 416 |
+
},
|
| 417 |
+
"pass_at_k": {
|
| 418 |
+
"1": {
|
| 419 |
+
"mean": 0.023875,
|
| 420 |
+
"all_task_mean": 0.023875,
|
| 421 |
+
"ci95": [
|
| 422 |
+
0.01578046875,
|
| 423 |
+
0.03222187499999998
|
| 424 |
+
],
|
| 425 |
+
"eligible_tasks": 500,
|
| 426 |
+
"total_tasks": 500,
|
| 427 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 428 |
+
"bootstrap_valid_replicates": 2000
|
| 429 |
+
},
|
| 430 |
+
"4": {
|
| 431 |
+
"mean": 0.011382711968975851,
|
| 432 |
+
"all_task_mean": 0.011382711968975851,
|
| 433 |
+
"ci95": [
|
| 434 |
+
0.0007472461188335719,
|
| 435 |
+
0.021702632693082518
|
| 436 |
+
],
|
| 437 |
+
"eligible_tasks": 500,
|
| 438 |
+
"total_tasks": 500,
|
| 439 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 440 |
+
"bootstrap_valid_replicates": 2000
|
| 441 |
+
},
|
| 442 |
+
"8": {
|
| 443 |
+
"mean": 0.007953911855287916,
|
| 444 |
+
"all_task_mean": 0.007953911855287916,
|
| 445 |
+
"ci95": [
|
| 446 |
+
-0.0032052019292176603,
|
| 447 |
+
0.02016559937775917
|
| 448 |
+
],
|
| 449 |
+
"eligible_tasks": 500,
|
| 450 |
+
"total_tasks": 500,
|
| 451 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 452 |
+
"bootstrap_valid_replicates": 2000
|
| 453 |
+
},
|
| 454 |
+
"16": {
|
| 455 |
+
"mean": 0.005326398453625931,
|
| 456 |
+
"all_task_mean": 0.005326398453625931,
|
| 457 |
+
"ci95": [
|
| 458 |
+
-0.008709203610150353,
|
| 459 |
+
0.01970317489123075
|
| 460 |
+
],
|
| 461 |
+
"eligible_tasks": 500,
|
| 462 |
+
"total_tasks": 500,
|
| 463 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 464 |
+
"bootstrap_valid_replicates": 2000
|
| 465 |
+
},
|
| 466 |
+
"64": {
|
| 467 |
+
"mean": 0.006,
|
| 468 |
+
"all_task_mean": 0.006,
|
| 469 |
+
"ci95": [
|
| 470 |
+
-0.018,
|
| 471 |
+
0.03
|
| 472 |
+
],
|
| 473 |
+
"eligible_tasks": 500,
|
| 474 |
+
"total_tasks": 500,
|
| 475 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 476 |
+
"bootstrap_valid_replicates": 2000
|
| 477 |
+
}
|
| 478 |
+
},
|
| 479 |
+
"implementation_coverage_at_k": {
|
| 480 |
+
"1": {
|
| 481 |
+
"mean": 0.023875000000000004,
|
| 482 |
+
"all_task_mean": 0.023875000000000004,
|
| 483 |
+
"ci95": [
|
| 484 |
+
0.01578046875,
|
| 485 |
+
0.03222187499999998
|
| 486 |
+
],
|
| 487 |
+
"eligible_tasks": 500,
|
| 488 |
+
"total_tasks": 500,
|
| 489 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 490 |
+
"bootstrap_valid_replicates": 2000
|
| 491 |
+
},
|
| 492 |
+
"4": {
|
| 493 |
+
"mean": -0.022815082722671284,
|
| 494 |
+
"all_task_mean": -0.022815082722671284,
|
| 495 |
+
"ci95": [
|
| 496 |
+
-0.05716847402168163,
|
| 497 |
+
0.012386312514164822
|
| 498 |
+
],
|
| 499 |
+
"eligible_tasks": 500,
|
| 500 |
+
"total_tasks": 500,
|
| 501 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 502 |
+
"bootstrap_valid_replicates": 2000
|
| 503 |
+
},
|
| 504 |
+
"8": {
|
| 505 |
+
"mean": -0.10593269048369632,
|
| 506 |
+
"all_task_mean": -0.10593269048369632,
|
| 507 |
+
"ci95": [
|
| 508 |
+
-0.17242767101744216,
|
| 509 |
+
-0.035957454487825634
|
| 510 |
+
],
|
| 511 |
+
"eligible_tasks": 500,
|
| 512 |
+
"total_tasks": 500,
|
| 513 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 514 |
+
"bootstrap_valid_replicates": 2000
|
| 515 |
+
},
|
| 516 |
+
"16": {
|
| 517 |
+
"mean": -0.28138206276553446,
|
| 518 |
+
"all_task_mean": -0.28138206276553446,
|
| 519 |
+
"ci95": [
|
| 520 |
+
-0.4102151369446903,
|
| 521 |
+
-0.146602173945259
|
| 522 |
+
],
|
| 523 |
+
"eligible_tasks": 500,
|
| 524 |
+
"total_tasks": 500,
|
| 525 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 526 |
+
"bootstrap_valid_replicates": 2000
|
| 527 |
+
},
|
| 528 |
+
"64": {
|
| 529 |
+
"mean": -1.292,
|
| 530 |
+
"all_task_mean": -1.292,
|
| 531 |
+
"ci95": [
|
| 532 |
+
-1.7860500000000001,
|
| 533 |
+
-0.78
|
| 534 |
+
],
|
| 535 |
+
"eligible_tasks": 500,
|
| 536 |
+
"total_tasks": 500,
|
| 537 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 538 |
+
"bootstrap_valid_replicates": 2000
|
| 539 |
+
}
|
| 540 |
+
},
|
| 541 |
+
"correct_matched_coverage_at_budgets": {
|
| 542 |
+
"4": {
|
| 543 |
+
"mean": -0.21621423592999497,
|
| 544 |
+
"all_task_mean": null,
|
| 545 |
+
"ci95": [
|
| 546 |
+
-0.26338146288134634,
|
| 547 |
+
-0.1712302020285684
|
| 548 |
+
],
|
| 549 |
+
"eligible_tasks": 311,
|
| 550 |
+
"total_tasks": 500,
|
| 551 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 552 |
+
"bootstrap_valid_replicates": 2000
|
| 553 |
+
},
|
| 554 |
+
"8": {
|
| 555 |
+
"mean": -0.5655161846473469,
|
| 556 |
+
"all_task_mean": null,
|
| 557 |
+
"ci95": [
|
| 558 |
+
-0.6710582647547939,
|
| 559 |
+
-0.4554695696815643
|
| 560 |
+
],
|
| 561 |
+
"eligible_tasks": 283,
|
| 562 |
+
"total_tasks": 500,
|
| 563 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 564 |
+
"bootstrap_valid_replicates": 2000
|
| 565 |
+
},
|
| 566 |
+
"16": {
|
| 567 |
+
"mean": -1.3422586793603524,
|
| 568 |
+
"all_task_mean": null,
|
| 569 |
+
"ci95": [
|
| 570 |
+
-1.5882434475943992,
|
| 571 |
+
-1.091048810798944
|
| 572 |
+
],
|
| 573 |
+
"eligible_tasks": 237,
|
| 574 |
+
"total_tasks": 500,
|
| 575 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 576 |
+
"bootstrap_valid_replicates": 2000
|
| 577 |
+
},
|
| 578 |
+
"64": {
|
| 579 |
+
"mean": -5.222222222222222,
|
| 580 |
+
"all_task_mean": null,
|
| 581 |
+
"ci95": [
|
| 582 |
+
-7.45125,
|
| 583 |
+
-3.0
|
| 584 |
+
],
|
| 585 |
+
"eligible_tasks": 18,
|
| 586 |
+
"total_tasks": 500,
|
| 587 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 588 |
+
"bootstrap_valid_replicates": 2000
|
| 589 |
+
}
|
| 590 |
+
},
|
| 591 |
+
"caveat": "Metric, sampling budgets and available declared evaluator/task harness identities were checked. Missing legacy provenance is explicitly unverified. Identical prompts, decoding protocol, and stable strategy annotation identities remain caller responsibilities. Partial-eligibility differences describe only the reported paired subset; these pointwise CIs do not establish an overall diversity gain."
|
| 592 |
+
},
|
| 593 |
+
"spectral_soft_minus_plain": {
|
| 594 |
+
"direction": "current_minus_previous",
|
| 595 |
+
"task_count": 500,
|
| 596 |
+
"evaluation_identity_check": "declared_provenance_equal",
|
| 597 |
+
"correctness": {
|
| 598 |
+
"metric": "task_macro_correct_fraction",
|
| 599 |
+
"margin": 0.01,
|
| 600 |
+
"delta": {
|
| 601 |
+
"mean": -0.01115625,
|
| 602 |
+
"all_task_mean": -0.01115625,
|
| 603 |
+
"ci95": [
|
| 604 |
+
-0.016094531250000002,
|
| 605 |
+
-0.00628125
|
| 606 |
+
],
|
| 607 |
+
"eligible_tasks": 500,
|
| 608 |
+
"total_tasks": 500,
|
| 609 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 610 |
+
"bootstrap_valid_replicates": 2000
|
| 611 |
+
},
|
| 612 |
+
"noninferior": false
|
| 613 |
+
},
|
| 614 |
+
"pass_at_k": {
|
| 615 |
+
"1": {
|
| 616 |
+
"mean": -0.01115625,
|
| 617 |
+
"all_task_mean": -0.01115625,
|
| 618 |
+
"ci95": [
|
| 619 |
+
-0.016094531250000002,
|
| 620 |
+
-0.00628125
|
| 621 |
+
],
|
| 622 |
+
"eligible_tasks": 500,
|
| 623 |
+
"total_tasks": 500,
|
| 624 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 625 |
+
"bootstrap_valid_replicates": 2000
|
| 626 |
+
},
|
| 627 |
+
"4": {
|
| 628 |
+
"mean": -0.003262993880788701,
|
| 629 |
+
"all_task_mean": -0.003262993880788701,
|
| 630 |
+
"ci95": [
|
| 631 |
+
-0.009283707049054424,
|
| 632 |
+
0.003047399256503232
|
| 633 |
+
],
|
| 634 |
+
"eligible_tasks": 500,
|
| 635 |
+
"total_tasks": 500,
|
| 636 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 637 |
+
"bootstrap_valid_replicates": 2000
|
| 638 |
+
},
|
| 639 |
+
"8": {
|
| 640 |
+
"mean": 0.0003559308731187021,
|
| 641 |
+
"all_task_mean": 0.0003559308731187021,
|
| 642 |
+
"ci95": [
|
| 643 |
+
-0.00660925510332175,
|
| 644 |
+
0.007311030447620167
|
| 645 |
+
],
|
| 646 |
+
"eligible_tasks": 500,
|
| 647 |
+
"total_tasks": 500,
|
| 648 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 649 |
+
"bootstrap_valid_replicates": 2000
|
| 650 |
+
},
|
| 651 |
+
"16": {
|
| 652 |
+
"mean": 0.005887105464296429,
|
| 653 |
+
"all_task_mean": 0.005887105464296429,
|
| 654 |
+
"ci95": [
|
| 655 |
+
-0.002448903637806618,
|
| 656 |
+
0.013853495529867147
|
| 657 |
+
],
|
| 658 |
+
"eligible_tasks": 500,
|
| 659 |
+
"total_tasks": 500,
|
| 660 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 661 |
+
"bootstrap_valid_replicates": 2000
|
| 662 |
+
},
|
| 663 |
+
"64": {
|
| 664 |
+
"mean": 0.022,
|
| 665 |
+
"all_task_mean": 0.022,
|
| 666 |
+
"ci95": [
|
| 667 |
+
0.004,
|
| 668 |
+
0.042
|
| 669 |
+
],
|
| 670 |
+
"eligible_tasks": 500,
|
| 671 |
+
"total_tasks": 500,
|
| 672 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 673 |
+
"bootstrap_valid_replicates": 2000
|
| 674 |
+
}
|
| 675 |
+
},
|
| 676 |
+
"implementation_coverage_at_k": {
|
| 677 |
+
"1": {
|
| 678 |
+
"mean": -0.011156250000000006,
|
| 679 |
+
"all_task_mean": -0.011156250000000006,
|
| 680 |
+
"ci95": [
|
| 681 |
+
-0.016094531250000006,
|
| 682 |
+
-0.006281250000000007
|
| 683 |
+
],
|
| 684 |
+
"eligible_tasks": 500,
|
| 685 |
+
"total_tasks": 500,
|
| 686 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 687 |
+
"bootstrap_valid_replicates": 2000
|
| 688 |
+
},
|
| 689 |
+
"4": {
|
| 690 |
+
"mean": 0.07761073757900834,
|
| 691 |
+
"all_task_mean": 0.07761073757900834,
|
| 692 |
+
"ci95": [
|
| 693 |
+
0.051705283406990515,
|
| 694 |
+
0.10339546725088768
|
| 695 |
+
],
|
| 696 |
+
"eligible_tasks": 500,
|
| 697 |
+
"total_tasks": 500,
|
| 698 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 699 |
+
"bootstrap_valid_replicates": 2000
|
| 700 |
+
},
|
| 701 |
+
"8": {
|
| 702 |
+
"mean": 0.23820672134995566,
|
| 703 |
+
"all_task_mean": 0.23820672134995566,
|
| 704 |
+
"ci95": [
|
| 705 |
+
0.18329618511067797,
|
| 706 |
+
0.29636921895754614
|
| 707 |
+
],
|
| 708 |
+
"eligible_tasks": 500,
|
| 709 |
+
"total_tasks": 500,
|
| 710 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 711 |
+
"bootstrap_valid_replicates": 2000
|
| 712 |
+
},
|
| 713 |
+
"16": {
|
| 714 |
+
"mean": 0.5990402364757417,
|
| 715 |
+
"all_task_mean": 0.5990402364757417,
|
| 716 |
+
"ci95": [
|
| 717 |
+
0.4869263191088721,
|
| 718 |
+
0.7231093071148845
|
| 719 |
+
],
|
| 720 |
+
"eligible_tasks": 500,
|
| 721 |
+
"total_tasks": 500,
|
| 722 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 723 |
+
"bootstrap_valid_replicates": 2000
|
| 724 |
+
},
|
| 725 |
+
"64": {
|
| 726 |
+
"mean": 3.004,
|
| 727 |
+
"all_task_mean": 3.004,
|
| 728 |
+
"ci95": [
|
| 729 |
+
2.568,
|
| 730 |
+
3.4940999999999995
|
| 731 |
+
],
|
| 732 |
+
"eligible_tasks": 500,
|
| 733 |
+
"total_tasks": 500,
|
| 734 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 735 |
+
"bootstrap_valid_replicates": 2000
|
| 736 |
+
}
|
| 737 |
+
},
|
| 738 |
+
"correct_matched_coverage_at_budgets": {
|
| 739 |
+
"4": {
|
| 740 |
+
"mean": 0.29542497837164544,
|
| 741 |
+
"all_task_mean": null,
|
| 742 |
+
"ci95": [
|
| 743 |
+
0.2488768199585551,
|
| 744 |
+
0.34316922020628526
|
| 745 |
+
],
|
| 746 |
+
"eligible_tasks": 316,
|
| 747 |
+
"total_tasks": 500,
|
| 748 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 749 |
+
"bootstrap_valid_replicates": 2000
|
| 750 |
+
},
|
| 751 |
+
"8": {
|
| 752 |
+
"mean": 0.7800955378748559,
|
| 753 |
+
"all_task_mean": null,
|
| 754 |
+
"ci95": [
|
| 755 |
+
0.6653322789038247,
|
| 756 |
+
0.8894675100835937
|
| 757 |
+
],
|
| 758 |
+
"eligible_tasks": 296,
|
| 759 |
+
"total_tasks": 500,
|
| 760 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 761 |
+
"bootstrap_valid_replicates": 2000
|
| 762 |
+
},
|
| 763 |
+
"16": {
|
| 764 |
+
"mean": 1.8157231276209642,
|
| 765 |
+
"all_task_mean": null,
|
| 766 |
+
"ci95": [
|
| 767 |
+
1.5818656638250101,
|
| 768 |
+
2.055080678055994
|
| 769 |
+
],
|
| 770 |
+
"eligible_tasks": 254,
|
| 771 |
+
"total_tasks": 500,
|
| 772 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 773 |
+
"bootstrap_valid_replicates": 2000
|
| 774 |
+
},
|
| 775 |
+
"64": {
|
| 776 |
+
"mean": 5.275862068965517,
|
| 777 |
+
"all_task_mean": null,
|
| 778 |
+
"ci95": [
|
| 779 |
+
3.5447214076246336,
|
| 780 |
+
7.150282258064515
|
| 781 |
+
],
|
| 782 |
+
"eligible_tasks": 29,
|
| 783 |
+
"total_tasks": 500,
|
| 784 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 785 |
+
"bootstrap_valid_replicates": 2000
|
| 786 |
+
}
|
| 787 |
+
},
|
| 788 |
+
"caveat": "Metric, sampling budgets and available declared evaluator/task harness identities were checked. Missing legacy provenance is explicitly unverified. Identical prompts, decoding protocol, and stable strategy annotation identities remain caller responsibilities. Partial-eligibility differences describe only the reported paired subset; these pointwise CIs do not establish an overall diversity gain."
|
| 789 |
+
},
|
| 790 |
+
"spectral_soft_minus_spd_hard": {
|
| 791 |
+
"direction": "current_minus_previous",
|
| 792 |
+
"task_count": 500,
|
| 793 |
+
"evaluation_identity_check": "declared_provenance_equal",
|
| 794 |
+
"correctness": {
|
| 795 |
+
"metric": "task_macro_correct_fraction",
|
| 796 |
+
"margin": 0.01,
|
| 797 |
+
"delta": {
|
| 798 |
+
"mean": -0.01434375,
|
| 799 |
+
"all_task_mean": -0.01434375,
|
| 800 |
+
"ci95": [
|
| 801 |
+
-0.01887578125,
|
| 802 |
+
-0.009780468750000004
|
| 803 |
+
],
|
| 804 |
+
"eligible_tasks": 500,
|
| 805 |
+
"total_tasks": 500,
|
| 806 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 807 |
+
"bootstrap_valid_replicates": 2000
|
| 808 |
+
},
|
| 809 |
+
"noninferior": false
|
| 810 |
+
},
|
| 811 |
+
"pass_at_k": {
|
| 812 |
+
"1": {
|
| 813 |
+
"mean": -0.01434375,
|
| 814 |
+
"all_task_mean": -0.01434375,
|
| 815 |
+
"ci95": [
|
| 816 |
+
-0.01887578125,
|
| 817 |
+
-0.009780468750000004
|
| 818 |
+
],
|
| 819 |
+
"eligible_tasks": 500,
|
| 820 |
+
"total_tasks": 500,
|
| 821 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 822 |
+
"bootstrap_valid_replicates": 2000
|
| 823 |
+
},
|
| 824 |
+
"4": {
|
| 825 |
+
"mean": -0.0038194990053133883,
|
| 826 |
+
"all_task_mean": -0.0038194990053133883,
|
| 827 |
+
"ci95": [
|
| 828 |
+
-0.009375225693132882,
|
| 829 |
+
0.00176260293117776
|
| 830 |
+
],
|
| 831 |
+
"eligible_tasks": 500,
|
| 832 |
+
"total_tasks": 500,
|
| 833 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 834 |
+
"bootstrap_valid_replicates": 2000
|
| 835 |
+
},
|
| 836 |
+
"8": {
|
| 837 |
+
"mean": 0.00015896448132888684,
|
| 838 |
+
"all_task_mean": 0.00015896448132888684,
|
| 839 |
+
"ci95": [
|
| 840 |
+
-0.006414460880441282,
|
| 841 |
+
0.006338201343282931
|
| 842 |
+
],
|
| 843 |
+
"eligible_tasks": 500,
|
| 844 |
+
"total_tasks": 500,
|
| 845 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 846 |
+
"bootstrap_valid_replicates": 2000
|
| 847 |
+
},
|
| 848 |
+
"16": {
|
| 849 |
+
"mean": 0.005789778461970222,
|
| 850 |
+
"all_task_mean": 0.005789778461970222,
|
| 851 |
+
"ci95": [
|
| 852 |
+
-0.0017539490758198626,
|
| 853 |
+
0.01354868460072776
|
| 854 |
+
],
|
| 855 |
+
"eligible_tasks": 500,
|
| 856 |
+
"total_tasks": 500,
|
| 857 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 858 |
+
"bootstrap_valid_replicates": 2000
|
| 859 |
+
},
|
| 860 |
+
"64": {
|
| 861 |
+
"mean": 0.024,
|
| 862 |
+
"all_task_mean": 0.024,
|
| 863 |
+
"ci95": [
|
| 864 |
+
0.006,
|
| 865 |
+
0.044
|
| 866 |
+
],
|
| 867 |
+
"eligible_tasks": 500,
|
| 868 |
+
"total_tasks": 500,
|
| 869 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 870 |
+
"bootstrap_valid_replicates": 2000
|
| 871 |
+
}
|
| 872 |
+
},
|
| 873 |
+
"implementation_coverage_at_k": {
|
| 874 |
+
"1": {
|
| 875 |
+
"mean": -0.014343750000000008,
|
| 876 |
+
"all_task_mean": -0.014343750000000008,
|
| 877 |
+
"ci95": [
|
| 878 |
+
-0.018875781250000008,
|
| 879 |
+
-0.009780468750000009
|
| 880 |
+
],
|
| 881 |
+
"eligible_tasks": 500,
|
| 882 |
+
"total_tasks": 500,
|
| 883 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 884 |
+
"bootstrap_valid_replicates": 2000
|
| 885 |
+
},
|
| 886 |
+
"4": {
|
| 887 |
+
"mean": 0.09105794364281937,
|
| 888 |
+
"all_task_mean": 0.09105794364281937,
|
| 889 |
+
"ci95": [
|
| 890 |
+
0.06601354867039361,
|
| 891 |
+
0.11605088018747954
|
| 892 |
+
],
|
| 893 |
+
"eligible_tasks": 500,
|
| 894 |
+
"total_tasks": 500,
|
| 895 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 896 |
+
"bootstrap_valid_replicates": 2000
|
| 897 |
+
},
|
| 898 |
+
"8": {
|
| 899 |
+
"mean": 0.2693537326077601,
|
| 900 |
+
"all_task_mean": 0.2693537326077601,
|
| 901 |
+
"ci95": [
|
| 902 |
+
0.21460300366072993,
|
| 903 |
+
0.3243236296632422
|
| 904 |
+
],
|
| 905 |
+
"eligible_tasks": 500,
|
| 906 |
+
"total_tasks": 500,
|
| 907 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 908 |
+
"bootstrap_valid_replicates": 2000
|
| 909 |
+
},
|
| 910 |
+
"16": {
|
| 911 |
+
"mean": 0.6548668763594453,
|
| 912 |
+
"all_task_mean": 0.6548668763594453,
|
| 913 |
+
"ci95": [
|
| 914 |
+
0.5437347177149345,
|
| 915 |
+
0.770010179781125
|
| 916 |
+
],
|
| 917 |
+
"eligible_tasks": 500,
|
| 918 |
+
"total_tasks": 500,
|
| 919 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 920 |
+
"bootstrap_valid_replicates": 2000
|
| 921 |
+
},
|
| 922 |
+
"64": {
|
| 923 |
+
"mean": 3.122,
|
| 924 |
+
"all_task_mean": 3.122,
|
| 925 |
+
"ci95": [
|
| 926 |
+
2.68595,
|
| 927 |
+
3.59605
|
| 928 |
+
],
|
| 929 |
+
"eligible_tasks": 500,
|
| 930 |
+
"total_tasks": 500,
|
| 931 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 932 |
+
"bootstrap_valid_replicates": 2000
|
| 933 |
+
}
|
| 934 |
+
},
|
| 935 |
+
"correct_matched_coverage_at_budgets": {
|
| 936 |
+
"4": {
|
| 937 |
+
"mean": 0.3343973321781965,
|
| 938 |
+
"all_task_mean": null,
|
| 939 |
+
"ci95": [
|
| 940 |
+
0.28955470590964033,
|
| 941 |
+
0.3787110897843099
|
| 942 |
+
],
|
| 943 |
+
"eligible_tasks": 318,
|
| 944 |
+
"total_tasks": 500,
|
| 945 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 946 |
+
"bootstrap_valid_replicates": 2000
|
| 947 |
+
},
|
| 948 |
+
"8": {
|
| 949 |
+
"mean": 0.8754529919459029,
|
| 950 |
+
"all_task_mean": null,
|
| 951 |
+
"ci95": [
|
| 952 |
+
0.7730941721429861,
|
| 953 |
+
0.9793856509650538
|
| 954 |
+
],
|
| 955 |
+
"eligible_tasks": 295,
|
| 956 |
+
"total_tasks": 500,
|
| 957 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 958 |
+
"bootstrap_valid_replicates": 2000
|
| 959 |
+
},
|
| 960 |
+
"16": {
|
| 961 |
+
"mean": 1.998443218444857,
|
| 962 |
+
"all_task_mean": null,
|
| 963 |
+
"ci95": [
|
| 964 |
+
1.7919974441259436,
|
| 965 |
+
2.2179767424598587
|
| 966 |
+
],
|
| 967 |
+
"eligible_tasks": 254,
|
| 968 |
+
"total_tasks": 500,
|
| 969 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 970 |
+
"bootstrap_valid_replicates": 2000
|
| 971 |
+
},
|
| 972 |
+
"64": {
|
| 973 |
+
"mean": 5.533333333333333,
|
| 974 |
+
"all_task_mean": null,
|
| 975 |
+
"ci95": [
|
| 976 |
+
3.928434065934066,
|
| 977 |
+
7.206932471264367
|
| 978 |
+
],
|
| 979 |
+
"eligible_tasks": 30,
|
| 980 |
+
"total_tasks": 500,
|
| 981 |
+
"population": "tasks_with_estimates_available_in_both_runs",
|
| 982 |
+
"bootstrap_valid_replicates": 2000
|
| 983 |
+
}
|
| 984 |
+
},
|
| 985 |
+
"caveat": "Metric, sampling budgets and available declared evaluator/task harness identities were checked. Missing legacy provenance is explicitly unverified. Identical prompts, decoding protocol, and stable strategy annotation identities remain caller responsibilities. Partial-eligibility differences describe only the reported paired subset; these pointwise CIs do not establish an overall diversity gain."
|
| 986 |
+
}
|
| 987 |
+
}
|
eval64/manifest.json
ADDED
|
@@ -0,0 +1,107 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"run": "retention_5round_single_seed_train16_eval16_b128_v1",
|
| 4 |
+
"training_seed": 43,
|
| 5 |
+
"train_tasks": 291,
|
| 6 |
+
"train_samples_per_task_per_round": 16,
|
| 7 |
+
"rounds": 5,
|
| 8 |
+
"eval_tasks": 500,
|
| 9 |
+
"eval_samples_per_task": 64,
|
| 10 |
+
"models": [
|
| 11 |
+
"base",
|
| 12 |
+
"plain",
|
| 13 |
+
"spd_hard",
|
| 14 |
+
"spectral_soft"
|
| 15 |
+
],
|
| 16 |
+
"ks": [
|
| 17 |
+
1,
|
| 18 |
+
4,
|
| 19 |
+
8,
|
| 20 |
+
16,
|
| 21 |
+
64
|
| 22 |
+
],
|
| 23 |
+
"correct_matched_budgets": [
|
| 24 |
+
4,
|
| 25 |
+
8,
|
| 26 |
+
16,
|
| 27 |
+
64
|
| 28 |
+
],
|
| 29 |
+
"evaluation_provenance_status": "identical task_local_v2 declarations across all four models",
|
| 30 |
+
"sources": [
|
| 31 |
+
{
|
| 32 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/base/evaluation.jsonl",
|
| 33 |
+
"bytes": 84523775,
|
| 34 |
+
"sha256": "007e9117e2850cd360e3cfbd4b7c25b24994d8a16ac39868a2b1a76be56b9de3"
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/base/evaluation.verified.jsonl",
|
| 38 |
+
"bytes": 117554831,
|
| 39 |
+
"sha256": "3491c161b0a8a3cf90879a3705e33a5fb38bd042c68327afbc89a3d8e393010d"
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/base/evaluation.metrics.json",
|
| 43 |
+
"bytes": 3342068,
|
| 44 |
+
"sha256": "a42d0319e011bd0ddafeff47d39ed23eedcd56505a45ddbbabbf38210254bc14"
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/plain/evaluation.jsonl",
|
| 48 |
+
"bytes": 68989331,
|
| 49 |
+
"sha256": "77b7070b4d4c40a4133451e1f0a26ea04327c13e808ba9da9816b924c5427e8c"
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/plain/evaluation.verified.jsonl",
|
| 53 |
+
"bytes": 98547609,
|
| 54 |
+
"sha256": "7f02e53bbbc28f18f0525ff9fb3d75f3b93e1a54ef00eb144fe9d6a1b81b5f4d"
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/plain/evaluation.metrics.json",
|
| 58 |
+
"bytes": 2980610,
|
| 59 |
+
"sha256": "c387c113f3051213632b8e60370bc214812aa53a425b37d47ee2168f47b956c5"
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/spd_hard/evaluation.jsonl",
|
| 63 |
+
"bytes": 65545873,
|
| 64 |
+
"sha256": "4e5e67df7e6c63146c7d32a06d0716abbeb40cbd194301f5fefac3146339b4f3"
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/spd_hard/evaluation.verified.jsonl",
|
| 68 |
+
"bytes": 94473226,
|
| 69 |
+
"sha256": "1e94017848bd71e0ca35ce28ee4b6633b25c417aa7038f453d2660b7f6cd018a"
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/spd_hard/evaluation.metrics.json",
|
| 73 |
+
"bytes": 2920343,
|
| 74 |
+
"sha256": "a6268c5e658c97a58c3ce74dba34aa1c7089c94283383ac28c58fa78c47d701b"
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/spectral_soft/evaluation.jsonl",
|
| 78 |
+
"bytes": 71794243,
|
| 79 |
+
"sha256": "1868edd31dc1ad383d67a09e91956f6b4cd3182162a2099186a9cf35731657ba"
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/spectral_soft/evaluation.verified.jsonl",
|
| 83 |
+
"bytes": 102663914,
|
| 84 |
+
"sha256": "3d7802860d0854f736653b91d390a83ff7a2c7d3e19cd3cd0ff41af555e8ac15"
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/spectral_soft/evaluation.metrics.json",
|
| 88 |
+
"bytes": 3188240,
|
| 89 |
+
"sha256": "1e3e64c77e9ab2c39c503d6afb5222501598dcdf3aeeccff1bc403e9291a983c"
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/summary.csv",
|
| 93 |
+
"bytes": 3390,
|
| 94 |
+
"sha256": "cb7eb3882257f6603025f5f2005d7d94734804ac75ef5ee7655cc01ccdd32d3f"
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/comparisons.json",
|
| 98 |
+
"bytes": 1558396,
|
| 99 |
+
"sha256": "dc35a5a934b487147b813410f4abb69953df25ad73721d72ee81fd4db736a598"
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"path": "runs/retention_5round_single_seed_train16_eval16_b128_v1/supplemental_eval64/REPORT.md",
|
| 103 |
+
"bytes": 3137,
|
| 104 |
+
"sha256": "72fba781e8aad9635bf06502249386a4f584c5da0ca9c22d622a6aadd4292799"
|
| 105 |
+
}
|
| 106 |
+
]
|
| 107 |
+
}
|
eval64/metrics_compact.json
ADDED
|
@@ -0,0 +1,766 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base": {
|
| 3 |
+
"task_count": 500,
|
| 4 |
+
"sample_count": 32000,
|
| 5 |
+
"correct_count": 12142,
|
| 6 |
+
"correct_fraction": {
|
| 7 |
+
"mean": 0.3794375,
|
| 8 |
+
"all_task_mean": 0.3794375,
|
| 9 |
+
"ci95": [
|
| 10 |
+
0.34740546875,
|
| 11 |
+
0.41365703125
|
| 12 |
+
],
|
| 13 |
+
"eligible_tasks": 500,
|
| 14 |
+
"total_tasks": 500,
|
| 15 |
+
"population": "tasks_with_available_estimates",
|
| 16 |
+
"bootstrap_valid_replicates": 2000
|
| 17 |
+
},
|
| 18 |
+
"pass_at_k": {
|
| 19 |
+
"1": {
|
| 20 |
+
"mean": 0.3794375,
|
| 21 |
+
"all_task_mean": 0.3794375,
|
| 22 |
+
"ci95": [
|
| 23 |
+
0.34740546875,
|
| 24 |
+
0.41365703125
|
| 25 |
+
],
|
| 26 |
+
"eligible_tasks": 500,
|
| 27 |
+
"total_tasks": 500,
|
| 28 |
+
"population": "tasks_with_available_estimates",
|
| 29 |
+
"bootstrap_valid_replicates": 2000
|
| 30 |
+
},
|
| 31 |
+
"4": {
|
| 32 |
+
"mean": 0.5507654491198912,
|
| 33 |
+
"all_task_mean": 0.5507654491198912,
|
| 34 |
+
"ci95": [
|
| 35 |
+
0.5132311306218681,
|
| 36 |
+
0.5890580618405479
|
| 37 |
+
],
|
| 38 |
+
"eligible_tasks": 500,
|
| 39 |
+
"total_tasks": 500,
|
| 40 |
+
"population": "tasks_with_available_estimates",
|
| 41 |
+
"bootstrap_valid_replicates": 2000
|
| 42 |
+
},
|
| 43 |
+
"8": {
|
| 44 |
+
"mean": 0.6109743277142735,
|
| 45 |
+
"all_task_mean": 0.6109743277142735,
|
| 46 |
+
"ci95": [
|
| 47 |
+
0.5724928416444336,
|
| 48 |
+
0.6497163629340588
|
| 49 |
+
],
|
| 50 |
+
"eligible_tasks": 500,
|
| 51 |
+
"total_tasks": 500,
|
| 52 |
+
"population": "tasks_with_available_estimates",
|
| 53 |
+
"bootstrap_valid_replicates": 2000
|
| 54 |
+
},
|
| 55 |
+
"16": {
|
| 56 |
+
"mean": 0.6569243830645763,
|
| 57 |
+
"all_task_mean": 0.6569243830645763,
|
| 58 |
+
"ci95": [
|
| 59 |
+
0.6184579345767386,
|
| 60 |
+
0.6952451294337093
|
| 61 |
+
],
|
| 62 |
+
"eligible_tasks": 500,
|
| 63 |
+
"total_tasks": 500,
|
| 64 |
+
"population": "tasks_with_available_estimates",
|
| 65 |
+
"bootstrap_valid_replicates": 2000
|
| 66 |
+
},
|
| 67 |
+
"64": {
|
| 68 |
+
"mean": 0.72,
|
| 69 |
+
"all_task_mean": 0.72,
|
| 70 |
+
"ci95": [
|
| 71 |
+
0.6819500000000001,
|
| 72 |
+
0.758
|
| 73 |
+
],
|
| 74 |
+
"eligible_tasks": 500,
|
| 75 |
+
"total_tasks": 500,
|
| 76 |
+
"population": "tasks_with_available_estimates",
|
| 77 |
+
"bootstrap_valid_replicates": 2000
|
| 78 |
+
}
|
| 79 |
+
},
|
| 80 |
+
"implementation_coverage_at_k": {
|
| 81 |
+
"1": {
|
| 82 |
+
"mean": 0.3794375,
|
| 83 |
+
"all_task_mean": 0.3794375,
|
| 84 |
+
"ci95": [
|
| 85 |
+
0.34740546875,
|
| 86 |
+
0.41365703125
|
| 87 |
+
],
|
| 88 |
+
"eligible_tasks": 500,
|
| 89 |
+
"total_tasks": 500,
|
| 90 |
+
"population": "tasks_with_available_estimates",
|
| 91 |
+
"bootstrap_valid_replicates": 2000
|
| 92 |
+
},
|
| 93 |
+
"4": {
|
| 94 |
+
"mean": 1.2665641226612274,
|
| 95 |
+
"all_task_mean": 1.2665641226612274,
|
| 96 |
+
"ci95": [
|
| 97 |
+
1.159415250025182,
|
| 98 |
+
1.3736470797606455
|
| 99 |
+
],
|
| 100 |
+
"eligible_tasks": 500,
|
| 101 |
+
"total_tasks": 500,
|
| 102 |
+
"population": "tasks_with_available_estimates",
|
| 103 |
+
"bootstrap_valid_replicates": 2000
|
| 104 |
+
},
|
| 105 |
+
"8": {
|
| 106 |
+
"mean": 2.2670132030155092,
|
| 107 |
+
"all_task_mean": 2.2670132030155092,
|
| 108 |
+
"ci95": [
|
| 109 |
+
2.0738004189885935,
|
| 110 |
+
2.4590917365037095
|
| 111 |
+
],
|
| 112 |
+
"eligible_tasks": 500,
|
| 113 |
+
"total_tasks": 500,
|
| 114 |
+
"population": "tasks_with_available_estimates",
|
| 115 |
+
"bootstrap_valid_replicates": 2000
|
| 116 |
+
},
|
| 117 |
+
"16": {
|
| 118 |
+
"mean": 4.039637834881175,
|
| 119 |
+
"all_task_mean": 4.039637834881175,
|
| 120 |
+
"ci95": [
|
| 121 |
+
3.6929032849396326,
|
| 122 |
+
4.390863679872534
|
| 123 |
+
],
|
| 124 |
+
"eligible_tasks": 500,
|
| 125 |
+
"total_tasks": 500,
|
| 126 |
+
"population": "tasks_with_available_estimates",
|
| 127 |
+
"bootstrap_valid_replicates": 2000
|
| 128 |
+
},
|
| 129 |
+
"64": {
|
| 130 |
+
"mean": 12.802,
|
| 131 |
+
"all_task_mean": 12.802,
|
| 132 |
+
"ci95": [
|
| 133 |
+
11.69585,
|
| 134 |
+
13.9681
|
| 135 |
+
],
|
| 136 |
+
"eligible_tasks": 500,
|
| 137 |
+
"total_tasks": 500,
|
| 138 |
+
"population": "tasks_with_available_estimates",
|
| 139 |
+
"bootstrap_valid_replicates": 2000
|
| 140 |
+
}
|
| 141 |
+
},
|
| 142 |
+
"correct_matched_coverage_at_budgets": {
|
| 143 |
+
"4": {
|
| 144 |
+
"mean": 3.3583723736089284,
|
| 145 |
+
"all_task_mean": null,
|
| 146 |
+
"ci95": [
|
| 147 |
+
3.282182510720265,
|
| 148 |
+
3.439956670735758
|
| 149 |
+
],
|
| 150 |
+
"eligible_tasks": 322,
|
| 151 |
+
"total_tasks": 500,
|
| 152 |
+
"population": "tasks_with_complete_labels_and_at_least_4_correct_samples",
|
| 153 |
+
"bootstrap_valid_replicates": 2000
|
| 154 |
+
},
|
| 155 |
+
"8": {
|
| 156 |
+
"mean": 5.986525949447583,
|
| 157 |
+
"all_task_mean": null,
|
| 158 |
+
"ci95": [
|
| 159 |
+
5.779330460447849,
|
| 160 |
+
6.196882660452379
|
| 161 |
+
],
|
| 162 |
+
"eligible_tasks": 294,
|
| 163 |
+
"total_tasks": 500,
|
| 164 |
+
"population": "tasks_with_complete_labels_and_at_least_8_correct_samples",
|
| 165 |
+
"bootstrap_valid_replicates": 2000
|
| 166 |
+
},
|
| 167 |
+
"16": {
|
| 168 |
+
"mean": 10.42331543495907,
|
| 169 |
+
"all_task_mean": null,
|
| 170 |
+
"ci95": [
|
| 171 |
+
9.959521542093748,
|
| 172 |
+
10.910493297809975
|
| 173 |
+
],
|
| 174 |
+
"eligible_tasks": 251,
|
| 175 |
+
"total_tasks": 500,
|
| 176 |
+
"population": "tasks_with_complete_labels_and_at_least_16_correct_samples",
|
| 177 |
+
"bootstrap_valid_replicates": 2000
|
| 178 |
+
},
|
| 179 |
+
"64": {
|
| 180 |
+
"mean": 19.875,
|
| 181 |
+
"all_task_mean": null,
|
| 182 |
+
"ci95": [
|
| 183 |
+
15.0,
|
| 184 |
+
25.04092307692307
|
| 185 |
+
],
|
| 186 |
+
"eligible_tasks": 24,
|
| 187 |
+
"total_tasks": 500,
|
| 188 |
+
"population": "tasks_with_complete_labels_and_at_least_64_correct_samples",
|
| 189 |
+
"bootstrap_valid_replicates": 2000
|
| 190 |
+
}
|
| 191 |
+
}
|
| 192 |
+
},
|
| 193 |
+
"plain": {
|
| 194 |
+
"task_count": 500,
|
| 195 |
+
"sample_count": 32000,
|
| 196 |
+
"correct_count": 13263,
|
| 197 |
+
"correct_fraction": {
|
| 198 |
+
"mean": 0.41446875,
|
| 199 |
+
"all_task_mean": 0.41446875,
|
| 200 |
+
"ci95": [
|
| 201 |
+
0.38040546875,
|
| 202 |
+
0.45168984375
|
| 203 |
+
],
|
| 204 |
+
"eligible_tasks": 500,
|
| 205 |
+
"total_tasks": 500,
|
| 206 |
+
"population": "tasks_with_available_estimates",
|
| 207 |
+
"bootstrap_valid_replicates": 2000
|
| 208 |
+
},
|
| 209 |
+
"pass_at_k": {
|
| 210 |
+
"1": {
|
| 211 |
+
"mean": 0.41446875,
|
| 212 |
+
"all_task_mean": 0.41446875,
|
| 213 |
+
"ci95": [
|
| 214 |
+
0.38040546875,
|
| 215 |
+
0.45168984375
|
| 216 |
+
],
|
| 217 |
+
"eligible_tasks": 500,
|
| 218 |
+
"total_tasks": 500,
|
| 219 |
+
"population": "tasks_with_available_estimates",
|
| 220 |
+
"bootstrap_valid_replicates": 2000
|
| 221 |
+
},
|
| 222 |
+
"4": {
|
| 223 |
+
"mean": 0.5654111549696559,
|
| 224 |
+
"all_task_mean": 0.5654111549696559,
|
| 225 |
+
"ci95": [
|
| 226 |
+
0.5268074897383598,
|
| 227 |
+
0.6045970197961521
|
| 228 |
+
],
|
| 229 |
+
"eligible_tasks": 500,
|
| 230 |
+
"total_tasks": 500,
|
| 231 |
+
"population": "tasks_with_available_estimates",
|
| 232 |
+
"bootstrap_valid_replicates": 2000
|
| 233 |
+
},
|
| 234 |
+
"8": {
|
| 235 |
+
"mean": 0.6185723086964429,
|
| 236 |
+
"all_task_mean": 0.6185723086964429,
|
| 237 |
+
"ci95": [
|
| 238 |
+
0.5793926136329006,
|
| 239 |
+
0.65815805619536
|
| 240 |
+
],
|
| 241 |
+
"eligible_tasks": 500,
|
| 242 |
+
"total_tasks": 500,
|
| 243 |
+
"population": "tasks_with_available_estimates",
|
| 244 |
+
"bootstrap_valid_replicates": 2000
|
| 245 |
+
},
|
| 246 |
+
"16": {
|
| 247 |
+
"mean": 0.6563636760539059,
|
| 248 |
+
"all_task_mean": 0.6563636760539059,
|
| 249 |
+
"ci95": [
|
| 250 |
+
0.6166121006457762,
|
| 251 |
+
0.6960772759859236
|
| 252 |
+
],
|
| 253 |
+
"eligible_tasks": 500,
|
| 254 |
+
"total_tasks": 500,
|
| 255 |
+
"population": "tasks_with_available_estimates",
|
| 256 |
+
"bootstrap_valid_replicates": 2000
|
| 257 |
+
},
|
| 258 |
+
"64": {
|
| 259 |
+
"mean": 0.704,
|
| 260 |
+
"all_task_mean": 0.704,
|
| 261 |
+
"ci95": [
|
| 262 |
+
0.664,
|
| 263 |
+
0.744
|
| 264 |
+
],
|
| 265 |
+
"eligible_tasks": 500,
|
| 266 |
+
"total_tasks": 500,
|
| 267 |
+
"population": "tasks_with_available_estimates",
|
| 268 |
+
"bootstrap_valid_replicates": 2000
|
| 269 |
+
}
|
| 270 |
+
},
|
| 271 |
+
"implementation_coverage_at_k": {
|
| 272 |
+
"1": {
|
| 273 |
+
"mean": 0.41446875,
|
| 274 |
+
"all_task_mean": 0.41446875,
|
| 275 |
+
"ci95": [
|
| 276 |
+
0.38040546875,
|
| 277 |
+
0.45168984375
|
| 278 |
+
],
|
| 279 |
+
"eligible_tasks": 500,
|
| 280 |
+
"total_tasks": 500,
|
| 281 |
+
"population": "tasks_with_available_estimates",
|
| 282 |
+
"bootstrap_valid_replicates": 2000
|
| 283 |
+
},
|
| 284 |
+
"4": {
|
| 285 |
+
"mean": 1.1661383023595477,
|
| 286 |
+
"all_task_mean": 1.1661383023595477,
|
| 287 |
+
"ci95": [
|
| 288 |
+
1.070903410028078,
|
| 289 |
+
1.2672731045868904
|
| 290 |
+
],
|
| 291 |
+
"eligible_tasks": 500,
|
| 292 |
+
"total_tasks": 500,
|
| 293 |
+
"population": "tasks_with_available_estimates",
|
| 294 |
+
"bootstrap_valid_replicates": 2000
|
| 295 |
+
},
|
| 296 |
+
"8": {
|
| 297 |
+
"mean": 1.9228737911818574,
|
| 298 |
+
"all_task_mean": 1.9228737911818574,
|
| 299 |
+
"ci95": [
|
| 300 |
+
1.7571056197779187,
|
| 301 |
+
2.101574059760503
|
| 302 |
+
],
|
| 303 |
+
"eligible_tasks": 500,
|
| 304 |
+
"total_tasks": 500,
|
| 305 |
+
"population": "tasks_with_available_estimates",
|
| 306 |
+
"bootstrap_valid_replicates": 2000
|
| 307 |
+
},
|
| 308 |
+
"16": {
|
| 309 |
+
"mean": 3.1592155356398983,
|
| 310 |
+
"all_task_mean": 3.1592155356398983,
|
| 311 |
+
"ci95": [
|
| 312 |
+
2.8673068380698647,
|
| 313 |
+
3.47591123889947
|
| 314 |
+
],
|
| 315 |
+
"eligible_tasks": 500,
|
| 316 |
+
"total_tasks": 500,
|
| 317 |
+
"population": "tasks_with_available_estimates",
|
| 318 |
+
"bootstrap_valid_replicates": 2000
|
| 319 |
+
},
|
| 320 |
+
"64": {
|
| 321 |
+
"mean": 8.506,
|
| 322 |
+
"all_task_mean": 8.506,
|
| 323 |
+
"ci95": [
|
| 324 |
+
7.60195,
|
| 325 |
+
9.46405
|
| 326 |
+
],
|
| 327 |
+
"eligible_tasks": 500,
|
| 328 |
+
"total_tasks": 500,
|
| 329 |
+
"population": "tasks_with_available_estimates",
|
| 330 |
+
"bootstrap_valid_replicates": 2000
|
| 331 |
+
}
|
| 332 |
+
},
|
| 333 |
+
"correct_matched_coverage_at_budgets": {
|
| 334 |
+
"4": {
|
| 335 |
+
"mean": 2.856476213813779,
|
| 336 |
+
"all_task_mean": null,
|
| 337 |
+
"ci95": [
|
| 338 |
+
2.7541549099452163,
|
| 339 |
+
2.9606601396926697
|
| 340 |
+
],
|
| 341 |
+
"eligible_tasks": 321,
|
| 342 |
+
"total_tasks": 500,
|
| 343 |
+
"population": "tasks_with_complete_labels_and_at_least_4_correct_samples",
|
| 344 |
+
"bootstrap_valid_replicates": 2000
|
| 345 |
+
},
|
| 346 |
+
"8": {
|
| 347 |
+
"mean": 4.690364473715527,
|
| 348 |
+
"all_task_mean": null,
|
| 349 |
+
"ci95": [
|
| 350 |
+
4.453332768614563,
|
| 351 |
+
4.9366147006484855
|
| 352 |
+
],
|
| 353 |
+
"eligible_tasks": 305,
|
| 354 |
+
"total_tasks": 500,
|
| 355 |
+
"population": "tasks_with_complete_labels_and_at_least_8_correct_samples",
|
| 356 |
+
"bootstrap_valid_replicates": 2000
|
| 357 |
+
},
|
| 358 |
+
"16": {
|
| 359 |
+
"mean": 7.3244981368337205,
|
| 360 |
+
"all_task_mean": null,
|
| 361 |
+
"ci95": [
|
| 362 |
+
6.814141127178505,
|
| 363 |
+
7.842568973120032
|
| 364 |
+
],
|
| 365 |
+
"eligible_tasks": 263,
|
| 366 |
+
"total_tasks": 500,
|
| 367 |
+
"population": "tasks_with_complete_labels_and_at_least_16_correct_samples",
|
| 368 |
+
"bootstrap_valid_replicates": 2000
|
| 369 |
+
},
|
| 370 |
+
"64": {
|
| 371 |
+
"mean": 8.686274509803921,
|
| 372 |
+
"all_task_mean": null,
|
| 373 |
+
"ci95": [
|
| 374 |
+
6.377226853883282,
|
| 375 |
+
11.162976249381494
|
| 376 |
+
],
|
| 377 |
+
"eligible_tasks": 51,
|
| 378 |
+
"total_tasks": 500,
|
| 379 |
+
"population": "tasks_with_complete_labels_and_at_least_64_correct_samples",
|
| 380 |
+
"bootstrap_valid_replicates": 2000
|
| 381 |
+
}
|
| 382 |
+
}
|
| 383 |
+
},
|
| 384 |
+
"spd_hard": {
|
| 385 |
+
"task_count": 500,
|
| 386 |
+
"sample_count": 32000,
|
| 387 |
+
"correct_count": 13365,
|
| 388 |
+
"correct_fraction": {
|
| 389 |
+
"mean": 0.41765625,
|
| 390 |
+
"all_task_mean": 0.41765625,
|
| 391 |
+
"ci95": [
|
| 392 |
+
0.384625,
|
| 393 |
+
0.4546578125
|
| 394 |
+
],
|
| 395 |
+
"eligible_tasks": 500,
|
| 396 |
+
"total_tasks": 500,
|
| 397 |
+
"population": "tasks_with_available_estimates",
|
| 398 |
+
"bootstrap_valid_replicates": 2000
|
| 399 |
+
},
|
| 400 |
+
"pass_at_k": {
|
| 401 |
+
"1": {
|
| 402 |
+
"mean": 0.41765625,
|
| 403 |
+
"all_task_mean": 0.41765625,
|
| 404 |
+
"ci95": [
|
| 405 |
+
0.384625,
|
| 406 |
+
0.4546578125
|
| 407 |
+
],
|
| 408 |
+
"eligible_tasks": 500,
|
| 409 |
+
"total_tasks": 500,
|
| 410 |
+
"population": "tasks_with_available_estimates",
|
| 411 |
+
"bootstrap_valid_replicates": 2000
|
| 412 |
+
},
|
| 413 |
+
"4": {
|
| 414 |
+
"mean": 0.5659676600941804,
|
| 415 |
+
"all_task_mean": 0.5659676600941804,
|
| 416 |
+
"ci95": [
|
| 417 |
+
0.527518317657576,
|
| 418 |
+
0.604875113633502
|
| 419 |
+
],
|
| 420 |
+
"eligible_tasks": 500,
|
| 421 |
+
"total_tasks": 500,
|
| 422 |
+
"population": "tasks_with_available_estimates",
|
| 423 |
+
"bootstrap_valid_replicates": 2000
|
| 424 |
+
},
|
| 425 |
+
"8": {
|
| 426 |
+
"mean": 0.6187692750882325,
|
| 427 |
+
"all_task_mean": 0.6187692750882325,
|
| 428 |
+
"ci95": [
|
| 429 |
+
0.5796125930854173,
|
| 430 |
+
0.6583837108361063
|
| 431 |
+
],
|
| 432 |
+
"eligible_tasks": 500,
|
| 433 |
+
"total_tasks": 500,
|
| 434 |
+
"population": "tasks_with_available_estimates",
|
| 435 |
+
"bootstrap_valid_replicates": 2000
|
| 436 |
+
},
|
| 437 |
+
"16": {
|
| 438 |
+
"mean": 0.656461003056232,
|
| 439 |
+
"all_task_mean": 0.656461003056232,
|
| 440 |
+
"ci95": [
|
| 441 |
+
0.616541219686637,
|
| 442 |
+
0.6960544516270536
|
| 443 |
+
],
|
| 444 |
+
"eligible_tasks": 500,
|
| 445 |
+
"total_tasks": 500,
|
| 446 |
+
"population": "tasks_with_available_estimates",
|
| 447 |
+
"bootstrap_valid_replicates": 2000
|
| 448 |
+
},
|
| 449 |
+
"64": {
|
| 450 |
+
"mean": 0.702,
|
| 451 |
+
"all_task_mean": 0.702,
|
| 452 |
+
"ci95": [
|
| 453 |
+
0.662,
|
| 454 |
+
0.7420499999999998
|
| 455 |
+
],
|
| 456 |
+
"eligible_tasks": 500,
|
| 457 |
+
"total_tasks": 500,
|
| 458 |
+
"population": "tasks_with_available_estimates",
|
| 459 |
+
"bootstrap_valid_replicates": 2000
|
| 460 |
+
}
|
| 461 |
+
},
|
| 462 |
+
"implementation_coverage_at_k": {
|
| 463 |
+
"1": {
|
| 464 |
+
"mean": 0.41765625,
|
| 465 |
+
"all_task_mean": 0.41765625,
|
| 466 |
+
"ci95": [
|
| 467 |
+
0.384625,
|
| 468 |
+
0.4546578125
|
| 469 |
+
],
|
| 470 |
+
"eligible_tasks": 500,
|
| 471 |
+
"total_tasks": 500,
|
| 472 |
+
"population": "tasks_with_available_estimates",
|
| 473 |
+
"bootstrap_valid_replicates": 2000
|
| 474 |
+
},
|
| 475 |
+
"4": {
|
| 476 |
+
"mean": 1.1526910962957369,
|
| 477 |
+
"all_task_mean": 1.1526910962957369,
|
| 478 |
+
"ci95": [
|
| 479 |
+
1.059633787473874,
|
| 480 |
+
1.2513327874361007
|
| 481 |
+
],
|
| 482 |
+
"eligible_tasks": 500,
|
| 483 |
+
"total_tasks": 500,
|
| 484 |
+
"population": "tasks_with_available_estimates",
|
| 485 |
+
"bootstrap_valid_replicates": 2000
|
| 486 |
+
},
|
| 487 |
+
"8": {
|
| 488 |
+
"mean": 1.8917267799240527,
|
| 489 |
+
"all_task_mean": 1.8917267799240527,
|
| 490 |
+
"ci95": [
|
| 491 |
+
1.7283757935135853,
|
| 492 |
+
2.062184461404335
|
| 493 |
+
],
|
| 494 |
+
"eligible_tasks": 500,
|
| 495 |
+
"total_tasks": 500,
|
| 496 |
+
"population": "tasks_with_available_estimates",
|
| 497 |
+
"bootstrap_valid_replicates": 2000
|
| 498 |
+
},
|
| 499 |
+
"16": {
|
| 500 |
+
"mean": 3.103388895756195,
|
| 501 |
+
"all_task_mean": 3.103388895756195,
|
| 502 |
+
"ci95": [
|
| 503 |
+
2.817451519989818,
|
| 504 |
+
3.40744240345712
|
| 505 |
+
],
|
| 506 |
+
"eligible_tasks": 500,
|
| 507 |
+
"total_tasks": 500,
|
| 508 |
+
"population": "tasks_with_available_estimates",
|
| 509 |
+
"bootstrap_valid_replicates": 2000
|
| 510 |
+
},
|
| 511 |
+
"64": {
|
| 512 |
+
"mean": 8.388,
|
| 513 |
+
"all_task_mean": 8.388,
|
| 514 |
+
"ci95": [
|
| 515 |
+
7.514,
|
| 516 |
+
9.322049999999999
|
| 517 |
+
],
|
| 518 |
+
"eligible_tasks": 500,
|
| 519 |
+
"total_tasks": 500,
|
| 520 |
+
"population": "tasks_with_available_estimates",
|
| 521 |
+
"bootstrap_valid_replicates": 2000
|
| 522 |
+
}
|
| 523 |
+
},
|
| 524 |
+
"correct_matched_coverage_at_budgets": {
|
| 525 |
+
"4": {
|
| 526 |
+
"mean": 2.8097969039470296,
|
| 527 |
+
"all_task_mean": null,
|
| 528 |
+
"ci95": [
|
| 529 |
+
2.702378107013863,
|
| 530 |
+
2.912624204119433
|
| 531 |
+
],
|
| 532 |
+
"eligible_tasks": 321,
|
| 533 |
+
"total_tasks": 500,
|
| 534 |
+
"population": "tasks_with_complete_labels_and_at_least_4_correct_samples",
|
| 535 |
+
"bootstrap_valid_replicates": 2000
|
| 536 |
+
},
|
| 537 |
+
"8": {
|
| 538 |
+
"mean": 4.589740759204707,
|
| 539 |
+
"all_task_mean": null,
|
| 540 |
+
"ci95": [
|
| 541 |
+
4.341571130697039,
|
| 542 |
+
4.8420976777276685
|
| 543 |
+
],
|
| 544 |
+
"eligible_tasks": 303,
|
| 545 |
+
"total_tasks": 500,
|
| 546 |
+
"population": "tasks_with_complete_labels_and_at_least_8_correct_samples",
|
| 547 |
+
"bootstrap_valid_replicates": 2000
|
| 548 |
+
},
|
| 549 |
+
"16": {
|
| 550 |
+
"mean": 7.136154971929802,
|
| 551 |
+
"all_task_mean": null,
|
| 552 |
+
"ci95": [
|
| 553 |
+
6.612424214688567,
|
| 554 |
+
7.64193275562665
|
| 555 |
+
],
|
| 556 |
+
"eligible_tasks": 264,
|
| 557 |
+
"total_tasks": 500,
|
| 558 |
+
"population": "tasks_with_complete_labels_and_at_least_16_correct_samples",
|
| 559 |
+
"bootstrap_valid_replicates": 2000
|
| 560 |
+
},
|
| 561 |
+
"64": {
|
| 562 |
+
"mean": 7.981132075471698,
|
| 563 |
+
"all_task_mean": null,
|
| 564 |
+
"ci95": [
|
| 565 |
+
5.9298245614035086,
|
| 566 |
+
10.380074468085107
|
| 567 |
+
],
|
| 568 |
+
"eligible_tasks": 53,
|
| 569 |
+
"total_tasks": 500,
|
| 570 |
+
"population": "tasks_with_complete_labels_and_at_least_64_correct_samples",
|
| 571 |
+
"bootstrap_valid_replicates": 2000
|
| 572 |
+
}
|
| 573 |
+
}
|
| 574 |
+
},
|
| 575 |
+
"spectral_soft": {
|
| 576 |
+
"task_count": 500,
|
| 577 |
+
"sample_count": 32000,
|
| 578 |
+
"correct_count": 12906,
|
| 579 |
+
"correct_fraction": {
|
| 580 |
+
"mean": 0.4033125,
|
| 581 |
+
"all_task_mean": 0.4033125,
|
| 582 |
+
"ci95": [
|
| 583 |
+
0.37115312500000003,
|
| 584 |
+
0.43831796874999995
|
| 585 |
+
],
|
| 586 |
+
"eligible_tasks": 500,
|
| 587 |
+
"total_tasks": 500,
|
| 588 |
+
"population": "tasks_with_available_estimates",
|
| 589 |
+
"bootstrap_valid_replicates": 2000
|
| 590 |
+
},
|
| 591 |
+
"pass_at_k": {
|
| 592 |
+
"1": {
|
| 593 |
+
"mean": 0.4033125,
|
| 594 |
+
"all_task_mean": 0.4033125,
|
| 595 |
+
"ci95": [
|
| 596 |
+
0.37115312500000003,
|
| 597 |
+
0.43831796874999995
|
| 598 |
+
],
|
| 599 |
+
"eligible_tasks": 500,
|
| 600 |
+
"total_tasks": 500,
|
| 601 |
+
"population": "tasks_with_available_estimates",
|
| 602 |
+
"bootstrap_valid_replicates": 2000
|
| 603 |
+
},
|
| 604 |
+
"4": {
|
| 605 |
+
"mean": 0.562148161088867,
|
| 606 |
+
"all_task_mean": 0.562148161088867,
|
| 607 |
+
"ci95": [
|
| 608 |
+
0.5239757778858503,
|
| 609 |
+
0.6009773591542646
|
| 610 |
+
],
|
| 611 |
+
"eligible_tasks": 500,
|
| 612 |
+
"total_tasks": 500,
|
| 613 |
+
"population": "tasks_with_available_estimates",
|
| 614 |
+
"bootstrap_valid_replicates": 2000
|
| 615 |
+
},
|
| 616 |
+
"8": {
|
| 617 |
+
"mean": 0.6189282395695613,
|
| 618 |
+
"all_task_mean": 0.6189282395695613,
|
| 619 |
+
"ci95": [
|
| 620 |
+
0.5799531764909873,
|
| 621 |
+
0.6581920368265824
|
| 622 |
+
],
|
| 623 |
+
"eligible_tasks": 500,
|
| 624 |
+
"total_tasks": 500,
|
| 625 |
+
"population": "tasks_with_available_estimates",
|
| 626 |
+
"bootstrap_valid_replicates": 2000
|
| 627 |
+
},
|
| 628 |
+
"16": {
|
| 629 |
+
"mean": 0.6622507815182023,
|
| 630 |
+
"all_task_mean": 0.6622507815182023,
|
| 631 |
+
"ci95": [
|
| 632 |
+
0.6233787417205545,
|
| 633 |
+
0.7018316378569717
|
| 634 |
+
],
|
| 635 |
+
"eligible_tasks": 500,
|
| 636 |
+
"total_tasks": 500,
|
| 637 |
+
"population": "tasks_with_available_estimates",
|
| 638 |
+
"bootstrap_valid_replicates": 2000
|
| 639 |
+
},
|
| 640 |
+
"64": {
|
| 641 |
+
"mean": 0.726,
|
| 642 |
+
"all_task_mean": 0.726,
|
| 643 |
+
"ci95": [
|
| 644 |
+
0.688,
|
| 645 |
+
0.766
|
| 646 |
+
],
|
| 647 |
+
"eligible_tasks": 500,
|
| 648 |
+
"total_tasks": 500,
|
| 649 |
+
"population": "tasks_with_available_estimates",
|
| 650 |
+
"bootstrap_valid_replicates": 2000
|
| 651 |
+
}
|
| 652 |
+
},
|
| 653 |
+
"implementation_coverage_at_k": {
|
| 654 |
+
"1": {
|
| 655 |
+
"mean": 0.4033125,
|
| 656 |
+
"all_task_mean": 0.4033125,
|
| 657 |
+
"ci95": [
|
| 658 |
+
0.37115312500000003,
|
| 659 |
+
0.43831796874999995
|
| 660 |
+
],
|
| 661 |
+
"eligible_tasks": 500,
|
| 662 |
+
"total_tasks": 500,
|
| 663 |
+
"population": "tasks_with_available_estimates",
|
| 664 |
+
"bootstrap_valid_replicates": 2000
|
| 665 |
+
},
|
| 666 |
+
"4": {
|
| 667 |
+
"mean": 1.243749039938556,
|
| 668 |
+
"all_task_mean": 1.243749039938556,
|
| 669 |
+
"ci95": [
|
| 670 |
+
1.1431083542658207,
|
| 671 |
+
1.348755115711012
|
| 672 |
+
],
|
| 673 |
+
"eligible_tasks": 500,
|
| 674 |
+
"total_tasks": 500,
|
| 675 |
+
"population": "tasks_with_available_estimates",
|
| 676 |
+
"bootstrap_valid_replicates": 2000
|
| 677 |
+
},
|
| 678 |
+
"8": {
|
| 679 |
+
"mean": 2.161080512531813,
|
| 680 |
+
"all_task_mean": 2.161080512531813,
|
| 681 |
+
"ci95": [
|
| 682 |
+
1.9763838957574165,
|
| 683 |
+
2.3468053037933396
|
| 684 |
+
],
|
| 685 |
+
"eligible_tasks": 500,
|
| 686 |
+
"total_tasks": 500,
|
| 687 |
+
"population": "tasks_with_available_estimates",
|
| 688 |
+
"bootstrap_valid_replicates": 2000
|
| 689 |
+
},
|
| 690 |
+
"16": {
|
| 691 |
+
"mean": 3.758255772115641,
|
| 692 |
+
"all_task_mean": 3.758255772115641,
|
| 693 |
+
"ci95": [
|
| 694 |
+
3.4251782302115035,
|
| 695 |
+
4.103086515045223
|
| 696 |
+
],
|
| 697 |
+
"eligible_tasks": 500,
|
| 698 |
+
"total_tasks": 500,
|
| 699 |
+
"population": "tasks_with_available_estimates",
|
| 700 |
+
"bootstrap_valid_replicates": 2000
|
| 701 |
+
},
|
| 702 |
+
"64": {
|
| 703 |
+
"mean": 11.51,
|
| 704 |
+
"all_task_mean": 11.51,
|
| 705 |
+
"ci95": [
|
| 706 |
+
10.424,
|
| 707 |
+
12.652399999999997
|
| 708 |
+
],
|
| 709 |
+
"eligible_tasks": 500,
|
| 710 |
+
"total_tasks": 500,
|
| 711 |
+
"population": "tasks_with_available_estimates",
|
| 712 |
+
"bootstrap_valid_replicates": 2000
|
| 713 |
+
}
|
| 714 |
+
},
|
| 715 |
+
"correct_matched_coverage_at_budgets": {
|
| 716 |
+
"4": {
|
| 717 |
+
"mean": 3.143291510215674,
|
| 718 |
+
"all_task_mean": null,
|
| 719 |
+
"ci95": [
|
| 720 |
+
3.052151432442491,
|
| 721 |
+
3.236868415837648
|
| 722 |
+
],
|
| 723 |
+
"eligible_tasks": 323,
|
| 724 |
+
"total_tasks": 500,
|
| 725 |
+
"population": "tasks_with_complete_labels_and_at_least_4_correct_samples",
|
| 726 |
+
"bootstrap_valid_replicates": 2000
|
| 727 |
+
},
|
| 728 |
+
"8": {
|
| 729 |
+
"mean": 5.451019508347272,
|
| 730 |
+
"all_task_mean": null,
|
| 731 |
+
"ci95": [
|
| 732 |
+
5.221458862962949,
|
| 733 |
+
5.691262607579812
|
| 734 |
+
],
|
| 735 |
+
"eligible_tasks": 302,
|
| 736 |
+
"total_tasks": 500,
|
| 737 |
+
"population": "tasks_with_complete_labels_and_at_least_8_correct_samples",
|
| 738 |
+
"bootstrap_valid_replicates": 2000
|
| 739 |
+
},
|
| 740 |
+
"16": {
|
| 741 |
+
"mean": 9.09058337320117,
|
| 742 |
+
"all_task_mean": null,
|
| 743 |
+
"ci95": [
|
| 744 |
+
8.580622492811118,
|
| 745 |
+
9.615960487699715
|
| 746 |
+
],
|
| 747 |
+
"eligible_tasks": 259,
|
| 748 |
+
"total_tasks": 500,
|
| 749 |
+
"population": "tasks_with_complete_labels_and_at_least_16_correct_samples",
|
| 750 |
+
"bootstrap_valid_replicates": 2000
|
| 751 |
+
},
|
| 752 |
+
"64": {
|
| 753 |
+
"mean": 12.176470588235293,
|
| 754 |
+
"all_task_mean": null,
|
| 755 |
+
"ci95": [
|
| 756 |
+
9.37483552631579,
|
| 757 |
+
15.0
|
| 758 |
+
],
|
| 759 |
+
"eligible_tasks": 34,
|
| 760 |
+
"total_tasks": 500,
|
| 761 |
+
"population": "tasks_with_complete_labels_and_at_least_64_correct_samples",
|
| 762 |
+
"bootstrap_valid_replicates": 2000
|
| 763 |
+
}
|
| 764 |
+
}
|
| 765 |
+
}
|
| 766 |
+
}
|
eval64/summary.csv
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model,tasks,samples,correct,pass_at_1,pass_at_1_ci_low,pass_at_1_ci_high,implementation_coverage_at_1,implementation_coverage_at_1_ci_low,implementation_coverage_at_1_ci_high,pass_at_4,pass_at_4_ci_low,pass_at_4_ci_high,implementation_coverage_at_4,implementation_coverage_at_4_ci_low,implementation_coverage_at_4_ci_high,pass_at_8,pass_at_8_ci_low,pass_at_8_ci_high,implementation_coverage_at_8,implementation_coverage_at_8_ci_low,implementation_coverage_at_8_ci_high,pass_at_16,pass_at_16_ci_low,pass_at_16_ci_high,implementation_coverage_at_16,implementation_coverage_at_16_ci_low,implementation_coverage_at_16_ci_high,pass_at_64,pass_at_64_ci_low,pass_at_64_ci_high,implementation_coverage_at_64,implementation_coverage_at_64_ci_low,implementation_coverage_at_64_ci_high,correct_matched_coverage_at_4,correct_matched_coverage_at_4_eligible_tasks,correct_matched_coverage_at_8,correct_matched_coverage_at_8_eligible_tasks,correct_matched_coverage_at_16,correct_matched_coverage_at_16_eligible_tasks,correct_matched_coverage_at_64,correct_matched_coverage_at_64_eligible_tasks
|
| 2 |
+
base,500,32000,12142,0.3794375,0.34740546875,0.41365703125,0.3794375,0.34740546875,0.41365703125,0.5507654491198912,0.5132311306218681,0.5890580618405479,1.2665641226612274,1.159415250025182,1.3736470797606455,0.6109743277142735,0.5724928416444336,0.6497163629340588,2.2670132030155092,2.0738004189885935,2.4590917365037095,0.6569243830645763,0.6184579345767386,0.6952451294337093,4.039637834881175,3.6929032849396326,4.390863679872534,0.72,0.6819500000000001,0.758,12.802,11.69585,13.9681,3.3583723736089284,322,5.986525949447583,294,10.42331543495907,251,19.875,24
|
| 3 |
+
plain,500,32000,13263,0.41446875,0.38040546875,0.45168984375,0.41446875,0.38040546875,0.45168984375,0.5654111549696559,0.5268074897383598,0.6045970197961521,1.1661383023595477,1.070903410028078,1.2672731045868904,0.6185723086964429,0.5793926136329006,0.65815805619536,1.9228737911818574,1.7571056197779187,2.101574059760503,0.6563636760539059,0.6166121006457762,0.6960772759859236,3.1592155356398983,2.8673068380698647,3.47591123889947,0.704,0.664,0.744,8.506,7.60195,9.46405,2.856476213813779,321,4.690364473715527,305,7.3244981368337205,263,8.686274509803921,51
|
| 4 |
+
spd_hard,500,32000,13365,0.41765625,0.384625,0.4546578125,0.41765625,0.384625,0.4546578125,0.5659676600941804,0.527518317657576,0.604875113633502,1.1526910962957369,1.059633787473874,1.2513327874361007,0.6187692750882325,0.5796125930854173,0.6583837108361063,1.8917267799240527,1.7283757935135853,2.062184461404335,0.656461003056232,0.616541219686637,0.6960544516270536,3.103388895756195,2.817451519989818,3.40744240345712,0.702,0.662,0.7420499999999998,8.388,7.514,9.322049999999999,2.8097969039470296,321,4.589740759204707,303,7.136154971929802,264,7.981132075471698,53
|
| 5 |
+
spectral_soft,500,32000,12906,0.4033125,0.37115312500000003,0.43831796874999995,0.4033125,0.37115312500000003,0.43831796874999995,0.562148161088867,0.5239757778858503,0.6009773591542646,1.243749039938556,1.1431083542658207,1.348755115711012,0.6189282395695613,0.5799531764909873,0.6581920368265824,2.161080512531813,1.9763838957574165,2.3468053037933396,0.6622507815182023,0.6233787417205545,0.7018316378569717,3.758255772115641,3.4251782302115035,4.103086515045223,0.726,0.688,0.766,11.51,10.424,12.652399999999997,3.143291510215674,323,5.451019508347272,302,9.09058337320117,259,12.176470588235293,34
|