List your company

Skill-based Attack Evaluation on SkillHarm Fixed-Payload Poisoning

86.3ASRGPT-5.4
25.04440.94756.8572.753Jun 1, 2026

Evaluation results

MethodLinks
GPT-5.4Harness=Codex2026.0686.390.72.8
GPT-5.5Harness=Codex2026.0681.485.64.6
Gemini 3 FlashHarness=Gemini CLI2026.0663.881.70.6
Qwen-3.6 27BHarness=OpenCode2026.0653.968.45.8
Sonnet 4.6Harness=Claude Code2026.0652.462.625.3
Opus 4.7Harness=Claude Code2026.0627.441.337.8