Obelum translator evals

Every published run, newest first. Each lives at its own address, /evals/<commit>/ (/evals/<commit>-single/ for single source), which does not change when a newer run is published. The newest of each: peer system · single source.

datesystemcommitmodelcases × runspassed
2026-10-14peer system907534bclaude-sonnet-5-553 × 3nodiff 108/159
2026-10-14single source4032790claude-sonnet-5-550 × 3nodiff 113/150
2026-10-12single source5b70239claude-sonnet-5-550 × 3hunks 150/150 · diff 147/150 · difflook 149/150
2026-10-12peer system5b70239claude-sonnet-5-553 × 3hunks 158/159 · diff 156/159 · difflook 159/159
2026-10-12single sourcee5a0bbfclaude-sonnet-5-549 × 3hunks 145/147 · diff 146/147 · nodiff 113/147
2026-10-12peer systeme5a0bbfclaude-sonnet-5-552 × 3hunks 155/156 · diff 153/156 · nodiff 114/156
2026-10-10single source6629f4aclaude-sonnet-5-549 × 3hunks 147/147 · nodiff 104/147
2026-10-10peer systemeb99223claude-sonnet-5-552 × 3hunks 153/156 · nodiff 113/156
2026-10-10peer system8522df6claude-sonnet-5-552 × 3hunks 155/156