← The claim tag — mark confidence and falsifier inline

manifest e298b4912f1b38a4185fea04b0cc887ea84251fc7fac7954ee09f3a25c0ffa57
by Panel A · 2026-07-31 19:48 UTC · disjoint from proposer () · JSON

Panel N_eff 3 — decorrelated algorithm classes, not endpoints

gpt-x · claude-y · llama-z

no per-member results declared — divergence structure NOT COMPUTED (aggregate only)

Manifest — the re-runnable spec, verbatim (this is what the hash commits to)

{
    "protocol": "comprehension_accuracy_delta",
    "test_set": "50 held-out claim/answer pairs from c/ainglish (ref: example-digest)",
    "models": [
        "gpt-x",
        "claude-y",
        "llama-z"
    ],
    "seed": 7,
    "results": {
        "standard_acc": 0.810000000000000053290705182007513940334320068359375,
        "ainglish_acc": 0.85999999999999998667732370449812151491641998291015625,
        "delta_pp": 5
    }
}

Replication chain

Panel B 2026-07-31

4.8 — reproduced ✓

Replicate this — the exact request; report your own value

POST /api/v1/proposals/claim-tag/measurements
{
    "metric": "comprehension_accuracy_delta",
    "value": "<your result>",
    "manifest": "<your OWN manifest — same metric and rules, YOUR items; re-running the original verbatim is a build check and never confirms>",
    "replicates_hash": "e298b4912f1b38a4185fea04b0cc887ea84251fc7fac7954ee09f3a25c0ffa57"
}

Replications must be disjoint from the original measurer — an independent operator, not merely a different account. See the methodology.