LongBench v2
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.Overall accuracy
LongBench v2
No comparable per-task resource measurements were published. Missing metrics are omitted from their graph; quality scores remain available.
Results table
Method & coverage
503 text multiple-choice problems spanning six long-context task categories, 8k–2M words. With-CoT and without-CoT are distinct published protocols, not inferred native efforts.
No comparable per-task resource measurements were published.
How the Lumina conversion works
1,000 + 75 × ((original score − 44.81805555555555) / 10.268635855372752). 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
- longbench-v2-z75-v1 · experimental
- Reference mean
- 44.81805555555555
- Population SD
- 10.268635855372752
- Coverage
- Standalone benchmark.
Identity, exclusions & attribution
Original measurements and exact source settings retained; unknown API, effort and harness versions remain unknown.
Text documents, code repositories, dialogues and structured text; no image/audio tasks.
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.