SWE-fficiency
Lumina VerifiedLumina checked the published sources, versions, model labels, score extraction and conversion. External evaluations were run by their publishers; Lumina has not independently rerun or reproduced them.Expert speedup ratio
SWE-fficiency
No comparable per-task resource measurements were published. Missing metrics are omitted from their graph; quality scores remain available.
Results table
Method & coverage
498 repository-level optimization tasks across nine Python repositories. Primary Speedup Ratio = model speedup / expert speedup; correctness is preserved separately, not blended by Lumina.
No comparable per-task resource measurements were published.
How the Lumina conversion works
1,000 + 75 × ((original score − 0.08901764705882353) / 0.05495363002829929). The result is rounded once. Scores may tie.
Frozen initial observed text-only cohort; one mean per exact publisher release/API, balancing repeated settings/routes within that release. Original score and protocol remain separate. No inferred variants or future refitting.
- Calibration
- swe-fficiency-z75-v1 · experimental
- Reference mean
- 0.08901764705882353
- Population SD
- 0.05495363002829929
- Coverage
- Standalone benchmark.
Identity, exclusions & attribution
Original measurements and exact source settings retained; unknown API, effort and harness versions remain unknown.
Text repositories, workloads, profiling and terminal-based correctness checks.
Standalone text benchmark; no new combined index. Source identity/effort/protocol qualifications remain in Data and CSV.
All published reasoning settings, including Claude Max, are included by default. Display labels use Low and xHigh; source labels remain in the retained records.
Publisher terms. Scores and resource figures are attributed to the original publisher.