Jason Weston's team introduces HackBench for detecting and localizing reward hacking in autoresearch agents

Mon, Oct 26 21:25 UTC+8

Sources1

@jasewestonXMon, Oct 26 21:25 UTC+8Introducing *HackBench* 🤡💀⏱️📊 ⏲️ 💪🔧 🔥 - Detecting & localizing reward hacking in autoresearch agents - Frontier models fail - Dataset on HuggingFace - Recipe to collect natural & synthetic hacks Blog post: facebookresearch.github.io/RAM/blogs/hack… 🧵1/5View original · Mon, Oct 26 21:25 UTC+8

Headline by dalao.ai. The quoted post is the record; its author keeps authorship.