From 5d7dfab2fae40fa54a26c4dc04bbf7cce7c0a5aa Mon Sep 17 00:00:00 2001 From: m3taversal Date: Thu, 19 Mar 2026 15:56:54 +0000 Subject: [PATCH] =?UTF-8?q?theseus:=20fix=2060%=20statistic=20precision=20?= =?UTF-8?q?=E2=80=94=20make=20conditional=20explicit?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Leo flagged: body text still read as unconditional probability. Now explicitly conditional: "when both err, ~60% of those errors are shared." Pentagon-Agent: Theseus <24DE7DA0-E4D5-4023-B1A2-3F736AFF4EEE> --- ...ror distributions that no same-family model can replicate.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/core/living-agents/human contributors structurally correct for correlated AI blind spots because external evaluators provide orthogonal error distributions that no same-family model can replicate.md b/core/living-agents/human contributors structurally correct for correlated AI blind spots because external evaluators provide orthogonal error distributions that no same-family model can replicate.md index 785f2424f..e41b9b654 100644 --- a/core/living-agents/human contributors structurally correct for correlated AI blind spots because external evaluators provide orthogonal error distributions that no same-family model can replicate.md +++ b/core/living-agents/human contributors structurally correct for correlated AI blind spots because external evaluators provide orthogonal error distributions that no same-family model can replicate.md @@ -31,7 +31,7 @@ Kim et al. (ICML 2025, "Correlated Errors in Large Language Models") evaluated 3 - Error correlation is highest for models sharing the **same base architecture** - As models get more accurate, their errors **converge** — the better they get, the more their mistakes overlap -This means our existing claim — [[all agents running the same model family creates correlated blind spots that adversarial review cannot catch because the evaluator shares the proposers training biases]] — is now empirically confirmed at scale. When a proposer agent makes an error, there is a ~60% chance that an evaluator agent from the same model family makes the same error — meaning roughly 6 out of 10 shared errors pass through review undetected. +This means our existing claim — [[all agents running the same model family creates correlated blind spots that adversarial review cannot catch because the evaluator shares the proposers training biases]] — is now empirically confirmed at scale. When both a proposer and evaluator from the same family err, ~60% of those errors are shared — meaning the evaluator cannot catch them because it makes the same mistake. The errors that slip through review are precisely the ones where shared training produces shared blind spots. ## Same-family evaluation has a structural self-preference bias