← All incidents

Drift incident · Raven Alpha (behaviour fingerprint)

gpt-6-astra

Model page
● OpenConfidence: Medium
Responses to fixed prompts no longer match this model’s own history (1344 recent vs 6300 baseline responses). Biggest shifts: answers on “Name a famous scientist. Reply with only the name.”; answers on “Give one tip for sleeping better, in one sentence.”; wording on “Give one tip for sleeping better, in one sentence.”
Nov 4change began ≈ 9:14 PM UTC
Nov 6incident opened 4:22 PM UTC
—still open
6.7e-4adjusted p

What this means

Raven Alpha compares how the model answers a fixed set of open-ended prompts with its own history. A persistent, significant shift means the model, its configuration or its serving path changed, even when every scored test still passes.

Opened Nov 6, 2026, 4:22 PM UTC UTC. An incident resolves on its own when the series returns to its baseline. Methodology.

Share

Permanent link: https://martingale-bench.vercel.app/incidents/inc-gpt-6-astra-fingerprint-2026-11-06T16

Incident id inc-gpt-6-astra-fingerprint-2026-11-06T16. Every public incident is also in the Atom feed and the JSON feed, and in the raw data. Methodology v0.2.