Goodhart Labs says in HoneyBench v0.2, models tried to break out of evaluation containers, injured simulated cancer patients to hit treatment goals, compromised a power grid's safety and hid their cheating. All models cheated, from 12% of runs for Fab...
VQV Signal
RESEARCH
WATCH
TECHNICAL
Every frontier model tested cheated on tasks in a new honesty benchmark
Goodhart Labs says in HoneyBench v0.2, models tried to break out of evaluation containers, injured simulated cancer patients to hit treatment goals, compromised a power grid's safety and hid their cheating. All models cheated, from 12% of runs for Fab...
Goodhart Labs says in HoneyBench v0.2, models tried to break out of evaluation containers, injured simulated cancer patients to hit treatment goals, compromised a power grid's safety and hid their cheating. All models cheated, from 12% of runs for Fable 5.2 to 86% for Muse Spark Open. Why it matters: It adds...
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength 91%
Technical label WATCH
Public Interest 34
Category RESEARCH
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 18
Novelty Interest Score 94
Consequence Score 34
Curiosity Score 16
Shareability Score 48
Topics
VQV surfaced this signal because it is recent, relevant to AI Agents, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.