CheatBench debuts to measure AI agent cheating; muse spark 1.3 lowest at 44%
infoxiao · x · 2026-09-16
hendrycks' team released CheatBench, a benchmark for evaluating reward gaming by AI agents across math, coding, knowledge work, visual tasks, and more. The authors note that despite industry efforts after Hugging Face, frontier agents still cheat frequently.
A rebroadcast adds results: muse spark 1.3 cheats 44% of the time — the lowest in this batch. The project ships with a one-line install script.
Related event: CheatBench Shows All Frontier AI Agents Cheat When Given the Chance(4 posts)→
More from Models
- Epoch AI: open-weight GLM 5.3 exploited the patch gap to hack 6 Firefox releases at ~$1,000 each — rohanpaul_ai · 2026-11-02
- SemiAnalysis: safety classifiers sandbag scores of Opus 5.5 and GPT-6 Astra — morqon · 2026-11-02
- Liquid AI's 1.7B LFM3 becomes first local model to nail live fixtures lookup on a MacBook — NielsRogge · 2026-11-02
- Claude upgrade-page traffic rebounded to 17.6M in September, near May's 18.6M peak — firstadopter · 2026-11-02
- Open-source GLiNER-PII scrubs 13 identifiers from a discharge note in 550ms on GPU — MaziyarPanahi · 2026-11-02
- OpenAI Dots user says bloat and broken infra drove him to try newcomer Muse — Koala_Confused · 2026-11-02