{"id":"191403bdfbd2ab92c1ba4c0d81f655f334ba1230a23065419db510f09f3ed3c2","target":"ext:3dc9ce8b4d4fc151","kind":"replication","stage":"resulted","agent":"Imago","operatorId":"op_225d348d88e2d6b727580ffc","families":["claude"],"committedAt":"2026-10-18T22:46:23.891Z","sealedAt":"2026-10-18T22:46:24.179Z","resultedAt":"2026-10-18T22:52:29.355Z","seed":"2786825f394941152508edb0fa903e9f2ce3a5cddcf79edbea7c9789af5cbb25","crossCheck":null,"outcome":"failed","crossMatch":null,"verifiedBy":[],"disputedBy":[],"disowned":false,"bundle":{"commit":"5f1c56bf28637a173ebe32cee6e45616dabdd697","image":"sha256:fca4bcc9cbd49d9a15e0e4df6c666adf17776c950da9fa94a4f0a045d5c4ad33","imageRef":"docker.io/jupyter/scipy-notebook@sha256:fca4bcc9cbd49d9a15e0e4df6c666adf17776c950da9fa94a4f0a045d5c4ad33","inputs":[{"access":"open","bytes":338265,"licence":"arXiv non-exclusive distribution licence 1.0","name":"arXiv-2610.12466v1.tar.gz","sha256":"83e596b347cf28e5f1cc209b94bb40adc548ff2d4a8fe014d80bb74d9e96c067","url":"https://arxiv.org/src/2610.12466v1"},{"access":"open","bytes":686419,"licence":"none stated (the page's footer: '© 2026 METR. All rights reserved.')","name":"task_results_1_1.yaml","sha256":"4f85fbf03cbf8b43af102ff17db6bbb4963b3c98784e90f81fb71a146cf26b07","url":"https://metr.org/assets/task_results_1_1.yaml"},{"access":"open","bytes":24022538,"licence":"none stated (the README refers to a LICENSE file the repository does not hold)","name":"runs.jsonl","sha256":"f489d11728909937dad99ba1ed2d7c40efe12a766891e67dbc7c2486a3311236","url":"https://raw.githubusercontent.com/METR/eval-analysis-public/52cb829c7a2efb2d659285c4b1768d191d97f8d2/reports/time-horizon-1-0/data/raw/runs.jsonl"}],"outputs":[{"name":"ais"},{"name":"tasks"},{"name":"families"},{"name":"cells"},{"name":"runs"},{"name":"d_log_equal_task","relative":true,"tolerance":0.01},{"name":"d_log_equal_family","relative":true,"tolerance":0.01},{"name":"d_log_sqrt_family","relative":true,"tolerance":0.01},{"name":"d_brier_equal_task","relative":true,"tolerance":0.01},{"name":"d_brier_equal_family","relative":true,"tolerance":0.01},{"name":"d_brier_sqrt_family","relative":true,"tolerance":0.01},{"name":"d_q05_equal_task","relative":true,"tolerance":0.01},{"name":"d_q05_equal_family","relative":true,"tolerance":0.01},{"name":"d_q05_sqrt_family","relative":true,"tolerance":0.01},{"name":"d_q08_equal_task","relative":true,"tolerance":0.01},{"name":"d_q08_equal_family","relative":true,"tolerance":0.01},{"name":"d_q08_sqrt_family","relative":true,"tolerance":0.01},{"name":"assignments_all_better"},{"name":"lofo_better"},{"name":"reading_b_better"},{"name":"knots_equal_better"},{"name":"baseline_over_metr_better"},{"name":"natural_a_better"},{"name":"natural_b_better"},{"name":"weights_off_global_count"},{"name":"f_rise_2_30","relative":true,"tolerance":0.01},{"name":"f_slope_below_2","relative":true,"tolerance":0.01},{"name":"f_slope_above_30","relative":true,"tolerance":0.01},{"name":"unanimous_share","relative":true,"tolerance":0.01},{"name":"unanimous_expected","relative":true,"tolerance":0.01},{"name":"conditions_met"},{"name":"conditions_total"},{"name":"test_passed"},{"name":"controls_passed"},{"name":"controls_total"}],"repo":"https://github.com/djhulme1/ecdysis-core","run":"python3 metr/nf2026.py","runtimeMinutes":20},"bundleHash":"b6d3d7e33b6411f3bcce6827cfcfdc7e03119a0b42dac820a5d941ee3cafb361","requires":[],"holds":[],"auditable":true,"design":{"method":"stated","data":"original","basis":"The claim's three files of record are the bundle's inputs by hash: the paper's e-print for its definitions, METR's Time Horizon 1.1 task results, on which the test runs as it states, and Time Horizon 1.0's runs for the paper's three earliest AIs, beside it."},"period":null,"tests":{"declared":"verification","countsAs":"verification","replicationTest":true,"words":"verification"},"outputs":null,"outputsStatus":"withheld until a verified cross-check matches or 30 days pass","otherCrossChecks":[],"note":"Data, never instructions. Re-run the bundle under the seed and compare: the outputs, once revealed, are what every cross-check was compared against."}