# Analyzer B - Version Leaderboard _Updated 2026-06-26T18:21:15. Benchmark: 4 confirmed pages - 355 E 86th (Cellar p2, 15th p13) + 953 Dean St (p4, p6). Detect-step time isolated (schedule/classify cached)._ ## Most accurate (by mean F1) | Version | Pages | F1 | Count acc | Tag acc | Prec | Rec | Box-IoU F1 | Detect s (sum) | Cost $ (sum) | pw | tw | dpi | tile | fpf | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | v26_consensus | 2 | 0.967 | 0.965 | 0.968 | 0.984 | 0.950 | 0.967 | 483.7 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v29_mergefix | 2 | 0.953 | 0.977 | 0.967 | 0.964 | 0.942 | 0.953 | 530.1 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v27_consensus3 | 2 | 0.944 | 0.972 | 0.968 | 0.938 | 0.950 | 0.944 | 2489.8 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v14_recall_max | 2 | 0.928 | 0.879 | 0.982 | 0.991 | 0.872 | 0.920 | 269.9 | $0.0000 | 1 | 12 | 300 | 1800 | True | | v17_ultra | 2 | 0.925 | 0.877 | 0.972 | 0.991 | 0.869 | 0.901 | 751.4 | $0.0000 | 1 | 16 | 300 | 1500 | True | | v30_runs1 | 4 | 0.907 | 0.844 | 0.965 | 0.885 | 0.947 | 0.789 | 782.4 | $1.5779 | 1 | 12 | 300 | 1800 | False | | v13_recall | 4 | 0.897 | 0.912 | 0.995 | 0.944 | 0.856 | 0.772 | 435.4 | $0.0000 | 1 | 12 | 300 | 1800 | True | | v19_maxrecall | 4 | 0.884 | 0.813 | 0.991 | 0.895 | 0.889 | 0.738 | 556.6 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v10_best | 4 | 0.881 | 0.807 | 0.952 | 0.862 | 0.924 | 0.740 | 1302.6 | $3.1337 | 1 | 12 | 300 | 1800 | False | | v28_adaptive | 4 | 0.881 | 0.716 | 0.956 | 0.857 | 0.948 | 0.763 | 1682.8 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v32_adaptive_prescan | 4 | 0.870 | 0.769 | 0.984 | 0.844 | 0.924 | 0.720 | 962.8 | $1.8632 | 1 | 12 | 300 | 1800 | False | | v25_max_direct | 2 | 0.862 | 0.771 | 0.963 | 1.000 | 0.771 | 0.862 | 306.9 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v47_or_qwen3vl_32b | 4 | 0.825 | 0.814 | 0.986 | 0.904 | 0.768 | 0.674 | 477.2 | $0.1881 | 1 | 8 | 300 | 1800 | False | | v73_qwen3vl_reason | 2 | 0.823 | 0.725 | 0.969 | 0.980 | 0.710 | 0.550 | 261.6 | $0.0874 | 1 | 8 | 300 | 1800 | False | | v41_or_flash_lite | 4 | 0.816 | 0.851 | 0.980 | 0.898 | 0.753 | 0.652 | 915.3 | $0.2121 | 1 | 8 | 300 | 1800 | False | | v80_or_glm5v | 4 | 0.807 | 0.683 | 0.986 | 0.814 | 0.832 | 0.707 | 29858.2 | $4.3354 | 1 | 8 | 300 | 1800 | False | | v48_or_gemini31_lite | 2 | 0.748 | 0.794 | 0.978 | 0.793 | 0.726 | 0.570 | 312.0 | $0.1505 | 1 | 8 | 300 | 1800 | False | | v03_tiles8 | 4 | 0.746 | 0.788 | 0.992 | 0.848 | 0.687 | 0.660 | 633.4 | $0.0000 | 1 | 8 | 300 | 3000 | True | | v22_or_g35 | 2 | 0.738 | 0.601 | 0.989 | 0.989 | 0.593 | 0.738 | 405.5 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v02_tiles4 | 2 | 0.691 | 0.906 | 1.000 | 0.700 | 0.684 | 0.531 | 941.7 | $0.0000 | 1 | 4 | 300 | 3000 | True | | v60_ensemble | 2 | 0.677 | 0.455 | 0.972 | 0.582 | 0.871 | 0.534 | 789.0 | $0.2316 | 1 | 8 | 300 | 1800 | False | | v40_or_gpt4o_mini | 2 | 0.665 | 0.570 | 0.969 | 0.922 | 0.522 | 0.665 | 321.9 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v46_or_gemma3_27b | 2 | 0.634 | 0.764 | 0.978 | 0.628 | 0.656 | 0.377 | 1187.6 | $0.0248 | 1 | 8 | 300 | 1800 | False | | v23_or_think | 2 | 0.611 | 0.445 | 0.986 | 1.000 | 0.445 | 0.582 | 609.6 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v21_or_maxrecall | 2 | 0.588 | 0.475 | 1.000 | 0.919 | 0.440 | 0.588 | 211.7 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v45_or_nova_lite | 2 | 0.565 | 0.379 | 0.853 | 0.456 | 0.802 | 0.428 | 1789.6 | $0.0671 | 1 | 8 | 300 | 1800 | False | | v50_ocr_easy | 4 | 0.438 | 0.294 | 0.722 | 1.000 | 0.294 | 0.356 | 17272.8 | $0.0000 | 1 | 4 | 300 | 3000 | True | | v70_sparse_qwen3vl | 2 | 0.411 | 0.259 | 1.000 | 1.000 | 0.259 | 0.263 | 256.4 | $0.0129 | 1 | 8 | 300 | 3000 | True | | v71_sparse_flashlite | 2 | 0.411 | 0.259 | 1.000 | 1.000 | 0.259 | 0.286 | 203.5 | $0.0169 | 1 | 8 | 300 | 3000 | True | | v44_or_nemotron_omni | 2 | 0.086 | 0.046 | 1.000 | 1.000 | 0.046 | 0.015 | 317.8 | $0.0000 | 1 | 6 | 300 | 1800 | False | ## Fastest (by total detect time over benchmark pages) | Version | Pages | F1 | Count acc | Tag acc | Prec | Rec | Box-IoU F1 | Detect s (sum) | Cost $ (sum) | pw | tw | dpi | tile | fpf | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | v71_sparse_flashlite | 2 | 0.411 | 0.259 | 1.000 | 1.000 | 0.259 | 0.286 | 203.5 | $0.0169 | 1 | 8 | 300 | 3000 | True | | v21_or_maxrecall | 2 | 0.588 | 0.475 | 1.000 | 0.919 | 0.440 | 0.588 | 211.7 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v70_sparse_qwen3vl | 2 | 0.411 | 0.259 | 1.000 | 1.000 | 0.259 | 0.263 | 256.4 | $0.0129 | 1 | 8 | 300 | 3000 | True | | v73_qwen3vl_reason | 2 | 0.823 | 0.725 | 0.969 | 0.980 | 0.710 | 0.550 | 261.6 | $0.0874 | 1 | 8 | 300 | 1800 | False | | v14_recall_max | 2 | 0.928 | 0.879 | 0.982 | 0.991 | 0.872 | 0.920 | 269.9 | $0.0000 | 1 | 12 | 300 | 1800 | True | | v25_max_direct | 2 | 0.862 | 0.771 | 0.963 | 1.000 | 0.771 | 0.862 | 306.9 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v48_or_gemini31_lite | 2 | 0.748 | 0.794 | 0.978 | 0.793 | 0.726 | 0.570 | 312.0 | $0.1505 | 1 | 8 | 300 | 1800 | False | | v44_or_nemotron_omni | 2 | 0.086 | 0.046 | 1.000 | 1.000 | 0.046 | 0.015 | 317.8 | $0.0000 | 1 | 6 | 300 | 1800 | False | | v40_or_gpt4o_mini | 2 | 0.665 | 0.570 | 0.969 | 0.922 | 0.522 | 0.665 | 321.9 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v22_or_g35 | 2 | 0.738 | 0.601 | 0.989 | 0.989 | 0.593 | 0.738 | 405.5 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v13_recall | 4 | 0.897 | 0.912 | 0.995 | 0.944 | 0.856 | 0.772 | 435.4 | $0.0000 | 1 | 12 | 300 | 1800 | True | | v47_or_qwen3vl_32b | 4 | 0.825 | 0.814 | 0.986 | 0.904 | 0.768 | 0.674 | 477.2 | $0.1881 | 1 | 8 | 300 | 1800 | False | | v26_consensus | 2 | 0.967 | 0.965 | 0.968 | 0.984 | 0.950 | 0.967 | 483.7 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v29_mergefix | 2 | 0.953 | 0.977 | 0.967 | 0.964 | 0.942 | 0.953 | 530.1 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v19_maxrecall | 4 | 0.884 | 0.813 | 0.991 | 0.895 | 0.889 | 0.738 | 556.6 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v23_or_think | 2 | 0.611 | 0.445 | 0.986 | 1.000 | 0.445 | 0.582 | 609.6 | $0.0000 | 1 | 8 | 300 | 1800 | False | | v03_tiles8 | 4 | 0.746 | 0.788 | 0.992 | 0.848 | 0.687 | 0.660 | 633.4 | $0.0000 | 1 | 8 | 300 | 3000 | True | | v17_ultra | 2 | 0.925 | 0.877 | 0.972 | 0.991 | 0.869 | 0.901 | 751.4 | $0.0000 | 1 | 16 | 300 | 1500 | True | | v30_runs1 | 4 | 0.907 | 0.844 | 0.965 | 0.885 | 0.947 | 0.789 | 782.4 | $1.5779 | 1 | 12 | 300 | 1800 | False | | v60_ensemble | 2 | 0.677 | 0.455 | 0.972 | 0.582 | 0.871 | 0.534 | 789.0 | $0.2316 | 1 | 8 | 300 | 1800 | False | | v41_or_flash_lite | 4 | 0.816 | 0.851 | 0.980 | 0.898 | 0.753 | 0.652 | 915.3 | $0.2121 | 1 | 8 | 300 | 1800 | False | | v02_tiles4 | 2 | 0.691 | 0.906 | 1.000 | 0.700 | 0.684 | 0.531 | 941.7 | $0.0000 | 1 | 4 | 300 | 3000 | True | | v32_adaptive_prescan | 4 | 0.870 | 0.769 | 0.984 | 0.844 | 0.924 | 0.720 | 962.8 | $1.8632 | 1 | 12 | 300 | 1800 | False | | v46_or_gemma3_27b | 2 | 0.634 | 0.764 | 0.978 | 0.628 | 0.656 | 0.377 | 1187.6 | $0.0248 | 1 | 8 | 300 | 1800 | False | | v10_best | 4 | 0.881 | 0.807 | 0.952 | 0.862 | 0.924 | 0.740 | 1302.6 | $3.1337 | 1 | 12 | 300 | 1800 | False | | v28_adaptive | 4 | 0.881 | 0.716 | 0.956 | 0.857 | 0.948 | 0.763 | 1682.8 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v45_or_nova_lite | 2 | 0.565 | 0.379 | 0.853 | 0.456 | 0.802 | 0.428 | 1789.6 | $0.0671 | 1 | 8 | 300 | 1800 | False | | v27_consensus3 | 2 | 0.944 | 0.972 | 0.968 | 0.938 | 0.950 | 0.944 | 2489.8 | $0.0000 | 1 | 12 | 300 | 1800 | False | | v50_ocr_easy | 4 | 0.438 | 0.294 | 0.722 | 1.000 | 0.294 | 0.356 | 17272.8 | $0.0000 | 1 | 4 | 300 | 3000 | True | | v80_or_glm5v | 4 | 0.807 | 0.683 | 0.986 | 0.814 | 0.832 | 0.707 | 29858.2 | $4.3354 | 1 | 8 | 300 | 1800 | False | ## Human-review workflow (fewest missed, then least to review) | Version | Pages | Recall | Found | of GT | MISSED | Review | Tag-review (yellow) | Bad green FP | Detect s | Cost $ | |---|---|---|---|---|---|---|---|---|---|---| | v26_consensus | 2 | 0.950 | 88 | 91 | 5 | 2 | 4 | 1 | 483.7 | $0.0000 | | v27_consensus3 | 2 | 0.950 | 91 | 91 | 5 | 4 | 4 | 1 | 2489.8 | $0.0000 | | v02_tiles4 | 2 | 0.684 | 22 | 22 | 6 | 0 | 0 | 6 | 941.7 | $0.0000 | | v29_mergefix | 2 | 0.942 | 88 | 91 | 6 | 3 | 4 | 1 | 530.1 | $0.0000 | | v30_runs1 | 4 | 0.947 | 115 | 113 | 6 | 10 | 5 | 3 | 782.4 | $1.5779 | | v10_best | 4 | 0.924 | 116 | 113 | 8 | 11 | 5 | 4 | 1302.6 | $3.1337 | | v28_adaptive | 4 | 0.948 | 117 | 113 | 8 | 13 | 5 | 3 | 1682.8 | $0.0000 | | v32_adaptive_prescan | 4 | 0.924 | 118 | 113 | 8 | 13 | 4 | 3 | 962.8 | $1.8632 | | v14_recall_max | 2 | 0.872 | 80 | 91 | 12 | 0 | 2 | 1 | 269.9 | $0.0000 | | v60_ensemble | 2 | 0.871 | 101 | 79 | 13 | 28 | 3 | 20 | 789.0 | $0.2316 | | v17_ultra | 2 | 0.869 | 78 | 91 | 14 | 0 | 3 | 1 | 751.4 | $0.0000 | | v19_maxrecall | 4 | 0.889 | 106 | 113 | 14 | 11 | 2 | 1 | 556.6 | $0.0000 | | v13_recall | 4 | 0.856 | 97 | 113 | 18 | 0 | 1 | 2 | 435.4 | $0.0000 | | v80_or_glm5v | 4 | 0.832 | 110 | 113 | 18 | 16 | 3 | 2 | 29858.2 | $4.3354 | | v73_qwen3vl_reason | 2 | 0.710 | 59 | 79 | 22 | 0 | 3 | 2 | 261.6 | $0.0874 | | v45_or_nova_lite | 2 | 0.802 | 115 | 79 | 22 | 61 | 5 | 1 | 1789.6 | $0.0671 | | v47_or_qwen3vl_32b | 4 | 0.768 | 95 | 113 | 24 | 2 | 3 | 4 | 477.2 | $0.1881 | | v48_or_gemini31_lite | 2 | 0.726 | 64 | 79 | 24 | 8 | 2 | 1 | 312.0 | $0.1505 | | v41_or_flash_lite | 4 | 0.753 | 95 | 113 | 25 | 8 | 4 | 2 | 915.3 | $0.2121 | | v46_or_gemma3_27b | 2 | 0.656 | 98 | 79 | 25 | 51 | 2 | 0 | 1187.6 | $0.0248 | | v25_max_direct | 2 | 0.771 | 64 | 91 | 27 | 2 | 3 | 0 | 306.9 | $0.0000 | | v22_or_g35 | 2 | 0.593 | 58 | 91 | 34 | 12 | 1 | 0 | 405.5 | $0.0000 | | v03_tiles8 | 4 | 0.687 | 76 | 113 | 45 | 0 | 1 | 8 | 633.4 | $0.0000 | | v40_or_gpt4o_mini | 2 | 0.522 | 49 | 91 | 45 | 3 | 2 | 1 | 321.9 | $0.0000 | | v21_or_maxrecall | 2 | 0.440 | 48 | 91 | 46 | 10 | 0 | 1 | 211.7 | $0.0000 | | v23_or_think | 2 | 0.445 | 44 | 91 | 47 | 0 | 1 | 0 | 609.6 | $0.0000 | | v70_sparse_qwen3vl | 2 | 0.259 | 22 | 79 | 57 | 0 | 0 | 0 | 256.4 | $0.0129 | | v71_sparse_flashlite | 2 | 0.259 | 22 | 79 | 57 | 0 | 0 | 0 | 203.5 | $0.0169 | | v44_or_nemotron_omni | 2 | 0.046 | 2 | 79 | 77 | 2 | 0 | 0 | 317.8 | $0.0000 | | v50_ocr_easy | 4 | 0.294 | 32 | 113 | 81 | 13 | 2 | 0 | 17272.8 | $0.0000 | _MISSED = units detected by nobody (the blind spot — a flagged-box review never sees these, so lower is critical). Review = boxes flagged for human verification (the only ones a person checks). Bad green FP = false positives auto-accepted as confirmed without review (want 0)._ ## Per-page detail (precision / recall / F1 per run) | Version | BP | Pg | GT | Pred | TP | FP | FN | Precision | Recall | F1 | Tag acc | Count acc | Detect s | |---|---|---|---|---|---|---|---|---|---|---|---|---|---| | v02_tiles4 | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 941.68 | | v02_tiles4 | 355 | 13 | 9 | 10 | 4 | 6 | 5 | 0.4 | 0.4444 | 0.4211 | 1.0 | 0.8889 | 941.68 | | v03_tiles8 | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 288.1 | | v03_tiles8 | 355 | 13 | 9 | 10 | 5 | 5 | 4 | 0.5 | 0.5556 | 0.5263 | 1.0 | 0.8889 | 288.1 | | v03_tiles8 | dean | 4 | 25 | 21 | 20 | 1 | 5 | 0.9524 | 0.8 | 0.8696 | 1.0 | 0.84 | 345.3 | | v03_tiles8 | dean | 6 | 66 | 33 | 31 | 2 | 35 | 0.9394 | 0.4697 | 0.6263 | 0.9677 | 0.5 | 345.3 | | v10_best | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 718.97 | | v10_best | 355 | 13 | 9 | 15 | 8 | 7 | 1 | 0.5333 | 0.8889 | 0.6667 | 0.875 | 0.3333 | 718.97 | | v10_best | dean | 4 | 25 | 25 | 24 | 1 | 1 | 0.96 | 0.96 | 0.96 | 1.0 | 1.0 | 583.66 | | v10_best | dean | 6 | 66 | 64 | 61 | 3 | 5 | 0.9531 | 0.9242 | 0.9385 | 0.9344 | 0.9697 | 583.66 | | v13_recall | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 190.94 | | v13_recall | 355 | 13 | 9 | 9 | 7 | 2 | 2 | 0.7778 | 0.7778 | 0.7778 | 1.0 | 1.0 | 190.94 | | v13_recall | dean | 4 | 25 | 23 | 23 | 0 | 2 | 1.0 | 0.92 | 0.9583 | 1.0 | 0.92 | 244.46 | | v13_recall | dean | 6 | 66 | 53 | 53 | 0 | 13 | 1.0 | 0.803 | 0.8908 | 0.9811 | 0.803 | 244.46 | | v14_recall_max | dean | 4 | 25 | 22 | 22 | 0 | 3 | 1.0 | 0.88 | 0.9362 | 1.0 | 0.88 | 269.87 | | v14_recall_max | dean | 6 | 66 | 58 | 57 | 1 | 9 | 0.9828 | 0.8636 | 0.9194 | 0.9649 | 0.8788 | 269.87 | | v17_ultra | dean | 4 | 25 | 23 | 23 | 0 | 2 | 1.0 | 0.92 | 0.9583 | 1.0 | 0.92 | 751.42 | | v17_ultra | dean | 6 | 66 | 55 | 54 | 1 | 12 | 0.9818 | 0.8182 | 0.8926 | 0.9444 | 0.8333 | 751.42 | | v19_maxrecall | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 288.82 | | v19_maxrecall | 355 | 13 | 9 | 13 | 8 | 5 | 1 | 0.6154 | 0.8889 | 0.7273 | 1.0 | 0.5556 | 288.82 | | v19_maxrecall | dean | 4 | 25 | 22 | 22 | 0 | 3 | 1.0 | 0.88 | 0.9362 | 1.0 | 0.88 | 267.77 | | v19_maxrecall | dean | 6 | 66 | 59 | 57 | 2 | 9 | 0.9661 | 0.8636 | 0.912 | 0.9649 | 0.8939 | 267.77 | | v21_or_maxrecall | dean | 4 | 25 | 9 | 8 | 1 | 17 | 0.8889 | 0.32 | 0.4706 | 1.0 | 0.36 | 211.72 | | v21_or_maxrecall | dean | 6 | 66 | 39 | 37 | 2 | 29 | 0.9487 | 0.5606 | 0.7048 | 1.0 | 0.5909 | 211.72 | | v22_or_g35 | dean | 4 | 25 | 13 | 13 | 0 | 12 | 1.0 | 0.52 | 0.6842 | 1.0 | 0.52 | 405.53 | | v22_or_g35 | dean | 6 | 66 | 45 | 44 | 1 | 22 | 0.9778 | 0.6667 | 0.7928 | 0.9773 | 0.6818 | 405.53 | | v23_or_think | dean | 4 | 25 | 9 | 9 | 0 | 16 | 1.0 | 0.36 | 0.5294 | 1.0 | 0.36 | 609.64 | | v23_or_think | dean | 6 | 66 | 35 | 35 | 0 | 31 | 1.0 | 0.5303 | 0.6931 | 0.9714 | 0.5303 | 609.64 | | v25_max_direct | dean | 4 | 25 | 23 | 23 | 0 | 2 | 1.0 | 0.92 | 0.9583 | 1.0 | 0.92 | 306.92 | | v25_max_direct | dean | 6 | 66 | 41 | 41 | 0 | 25 | 1.0 | 0.6212 | 0.7664 | 0.9268 | 0.6212 | 306.92 | | v26_consensus | dean | 4 | 25 | 24 | 24 | 0 | 1 | 1.0 | 0.96 | 0.9796 | 1.0 | 0.96 | 483.74 | | v26_consensus | dean | 6 | 66 | 64 | 62 | 2 | 4 | 0.9688 | 0.9394 | 0.9538 | 0.9355 | 0.9697 | 483.74 | | v27_consensus3 | dean | 4 | 25 | 26 | 24 | 2 | 1 | 0.9231 | 0.96 | 0.9412 | 1.0 | 0.96 | 2489.85 | | v27_consensus3 | dean | 6 | 66 | 65 | 62 | 3 | 4 | 0.9538 | 0.9394 | 0.9466 | 0.9355 | 0.9848 | 2489.85 | | v28_adaptive | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 599.28 | | v28_adaptive | 355 | 13 | 9 | 18 | 9 | 9 | 0 | 0.5 | 1.0 | 0.6667 | 0.8889 | 0.0 | 599.28 | | v28_adaptive | dean | 4 | 25 | 25 | 24 | 1 | 1 | 0.96 | 0.96 | 0.96 | 1.0 | 1.0 | 1083.57 | | v28_adaptive | dean | 6 | 66 | 62 | 60 | 2 | 6 | 0.9677 | 0.9091 | 0.9375 | 0.9333 | 0.9394 | 1083.57 | | v29_mergefix | dean | 4 | 25 | 25 | 24 | 1 | 1 | 0.96 | 0.96 | 0.96 | 1.0 | 1.0 | 530.1 | | v29_mergefix | dean | 6 | 66 | 63 | 61 | 2 | 5 | 0.9683 | 0.9242 | 0.9457 | 0.9344 | 0.9545 | 530.1 | | v30_runs1 | 355 | 2 | 13 | 13 | 13 | 0 | 0 | 1.0 | 1.0 | 1.0 | 0.9231 | 1.0 | 433.28 | | v30_runs1 | 355 | 13 | 9 | 14 | 8 | 6 | 1 | 0.5714 | 0.8889 | 0.6957 | 1.0 | 0.4444 | 433.28 | | v30_runs1 | dean | 4 | 25 | 24 | 24 | 0 | 1 | 1.0 | 0.96 | 0.9796 | 1.0 | 0.96 | 349.09 | | v30_runs1 | dean | 6 | 66 | 64 | 62 | 2 | 4 | 0.9688 | 0.9394 | 0.9538 | 0.9355 | 0.9697 | 349.09 | | v32_adaptive_prescan | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 513.92 | | v32_adaptive_prescan | 355 | 13 | 9 | 16 | 8 | 8 | 1 | 0.5 | 0.8889 | 0.64 | 1.0 | 0.2222 | 513.92 | | v32_adaptive_prescan | dean | 4 | 25 | 26 | 24 | 2 | 1 | 0.9231 | 0.96 | 0.9412 | 1.0 | 0.96 | 448.85 | | v32_adaptive_prescan | dean | 6 | 66 | 64 | 61 | 3 | 5 | 0.9531 | 0.9242 | 0.9385 | 0.9344 | 0.9697 | 448.85 | | v40_or_gpt4o_mini | dean | 4 | 25 | 16 | 14 | 2 | 11 | 0.875 | 0.56 | 0.6829 | 1.0 | 0.64 | 321.87 | | v40_or_gpt4o_mini | dean | 6 | 66 | 33 | 32 | 1 | 34 | 0.9697 | 0.4848 | 0.6465 | 0.9375 | 0.5 | 321.87 | | v41_or_flash_lite | 355 | 2 | 13 | 9 | 9 | 0 | 4 | 1.0 | 0.6923 | 0.8182 | 1.0 | 0.6923 | 420.59 | | v41_or_flash_lite | 355 | 13 | 9 | 9 | 6 | 3 | 3 | 0.6667 | 0.6667 | 0.6667 | 1.0 | 1.0 | 420.59 | | v41_or_flash_lite | dean | 4 | 25 | 22 | 22 | 0 | 3 | 1.0 | 0.88 | 0.9362 | 1.0 | 0.88 | 494.7 | | v41_or_flash_lite | dean | 6 | 66 | 55 | 51 | 4 | 15 | 0.9273 | 0.7727 | 0.843 | 0.9216 | 0.8333 | 494.7 | | v44_or_nemotron_omni | 355 | 2 | 13 | 1 | 1 | 0 | 12 | 1.0 | 0.0769 | 0.1429 | 1.0 | 0.0769 | 160.87 | | v44_or_nemotron_omni | dean | 6 | 66 | 1 | 1 | 0 | 65 | 1.0 | 0.0152 | 0.0299 | 1.0 | 0.0152 | 156.96 | | v45_or_nova_lite | 355 | 2 | 13 | 33 | 12 | 21 | 1 | 0.3636 | 0.9231 | 0.5217 | 0.75 | 0.0 | 1040.88 | | v45_or_nova_lite | dean | 6 | 66 | 82 | 45 | 37 | 21 | 0.5488 | 0.6818 | 0.6081 | 0.9556 | 0.7576 | 748.74 | | v46_or_gemma3_27b | 355 | 2 | 13 | 11 | 8 | 3 | 5 | 0.7273 | 0.6154 | 0.6667 | 1.0 | 0.8462 | 468.86 | | v46_or_gemma3_27b | dean | 6 | 66 | 87 | 46 | 41 | 20 | 0.5287 | 0.697 | 0.6013 | 0.9565 | 0.6818 | 718.73 | | v47_or_qwen3vl_32b | 355 | 2 | 13 | 9 | 9 | 0 | 4 | 1.0 | 0.6923 | 0.8182 | 1.0 | 0.6923 | 225.41 | | v47_or_qwen3vl_32b | 355 | 13 | 9 | 10 | 7 | 3 | 2 | 0.7 | 0.7778 | 0.7368 | 1.0 | 0.8889 | 225.41 | | v47_or_qwen3vl_32b | dean | 4 | 25 | 21 | 20 | 1 | 5 | 0.9524 | 0.8 | 0.8696 | 1.0 | 0.84 | 251.78 | | v47_or_qwen3vl_32b | dean | 6 | 66 | 55 | 53 | 2 | 13 | 0.9636 | 0.803 | 0.876 | 0.9434 | 0.8333 | 251.78 | | v48_or_gemini31_lite | 355 | 2 | 13 | 15 | 10 | 5 | 3 | 0.6667 | 0.7692 | 0.7143 | 1.0 | 0.8462 | 143.28 | | v48_or_gemini31_lite | dean | 6 | 66 | 49 | 45 | 4 | 21 | 0.9184 | 0.6818 | 0.7826 | 0.9556 | 0.7424 | 168.72 | | v50_ocr_easy | 355 | 2 | 13 | 6 | 6 | 0 | 7 | 1.0 | 0.4615 | 0.6316 | 1.0 | 0.4615 | 5913.04 | | v50_ocr_easy | 355 | 13 | 9 | 1 | 1 | 0 | 8 | 1.0 | 0.1111 | 0.2 | 0.0 | 0.1111 | 5913.04 | | v50_ocr_easy | dean | 4 | 25 | 9 | 9 | 0 | 16 | 1.0 | 0.36 | 0.5294 | 0.8889 | 0.36 | 11359.78 | | v50_ocr_easy | dean | 6 | 66 | 16 | 16 | 0 | 50 | 1.0 | 0.2424 | 0.3902 | 1.0 | 0.2424 | 11359.78 | | v60_ensemble | 355 | 2 | 13 | 29 | 12 | 17 | 1 | 0.4138 | 0.9231 | 0.5714 | 1.0 | 0.0 | 366.9 | | v60_ensemble | dean | 6 | 66 | 72 | 54 | 18 | 12 | 0.75 | 0.8182 | 0.7826 | 0.9444 | 0.9091 | 422.07 | | v70_sparse_qwen3vl | 355 | 2 | 13 | 3 | 3 | 0 | 10 | 1.0 | 0.2308 | 0.375 | 1.0 | 0.2308 | 135.0 | | v70_sparse_qwen3vl | dean | 6 | 66 | 19 | 19 | 0 | 47 | 1.0 | 0.2879 | 0.4471 | 1.0 | 0.2879 | 121.36 | | v71_sparse_flashlite | 355 | 2 | 13 | 3 | 3 | 0 | 10 | 1.0 | 0.2308 | 0.375 | 1.0 | 0.2308 | 78.05 | | v71_sparse_flashlite | dean | 6 | 66 | 19 | 19 | 0 | 47 | 1.0 | 0.2879 | 0.4471 | 1.0 | 0.2879 | 125.44 | | v73_qwen3vl_reason | 355 | 2 | 13 | 9 | 9 | 0 | 4 | 1.0 | 0.6923 | 0.8182 | 1.0 | 0.6923 | 105.96 | | v73_qwen3vl_reason | dean | 6 | 66 | 50 | 48 | 2 | 18 | 0.96 | 0.7273 | 0.8276 | 0.9375 | 0.7576 | 155.63 | | v80_or_glm5v | 355 | 2 | 13 | 12 | 12 | 0 | 1 | 1.0 | 0.9231 | 0.96 | 1.0 | 0.9231 | 28775.91 | | v80_or_glm5v | 355 | 13 | 9 | 18 | 6 | 12 | 3 | 0.3333 | 0.6667 | 0.4444 | 1.0 | 0.0 | 28775.91 | | v80_or_glm5v | dean | 4 | 25 | 24 | 23 | 1 | 2 | 0.9583 | 0.92 | 0.9388 | 1.0 | 0.96 | 1082.29 | | v80_or_glm5v | dean | 6 | 66 | 56 | 54 | 2 | 12 | 0.9643 | 0.8182 | 0.8852 | 0.9444 | 0.8485 | 1082.29 | _Detect s on a page row is the whole-blueprint invocation time (both its pages run together), not that single page._ _F1 (headline) = LOCATION match: a unit counts as found when a prediction's CENTER is within 0.7x the unit's size of a confirmed unit (tag-agnostic) — reflects the counting goal without penalising box tightness or a wrong tag. A located unit whose TAG is wrong is still counted as found but flagged YELLOW for tag review (see 'Tag-review' in the workflow table); Tag acc = fraction of located units tagged correctly. Box-IoU F1 = strict box-overlap F1 (IoU>=0.3), informational. Count acc = 1-|pred-gt|/gt. Detect s = sum over benchmarked pages._