CONSONANCE.for your information
Monday, 5 October 2026frenvi

Being discussed

01 — evaluation 3 VOICES

Evaluating safety refusal rates on coding agent benchmarks

Artificial Analysis introduced safety refusal reporting in version 1.5 of its Coding Agent Index to help explain model behavioral differences. Data showed distinct variations in safety refusal rates among evaluated models during coding tasks.

3 independent accounts 5 posts 1,583 interactions
devin
@ArtificialAnlys their topics on X ↗
@Hesamation their topics on X ↗
@NVIDIAAI their topics on X ↗
@0x15f their topics on X ↗
↑