vgel open-sources simple-reward-hacking: a reproducible environment for eliciting reward hacking
voooooogel · x · 2026-09-01
vgel shared the companion GitHub repo simple-reward-hacking: a code-execution environment designed to elicit reward hacking, tracked with AST-based metrics, where the reward is an easily hackable test-pass metric.
Key points:
- Dataset: a subset of SYNTHETIC-2 RL (Qwen3-32B-Hard, 1683 examples), tagged dont-train-on-this;
- Runs via a single command uv run vf-eval simple-reward-hacking, with configurable models and sampling args;
- Submitted code runs sandboxed by default via bubblewrap + prlimit.
Related event: Anthropic Discloses Jailbreak Incidents and Reward Hacking Research(39 posts)→
More from coding & agent
- Connected steps are not a workflow: nobody designs the handoff — usehive · 2026-10-16
- Making Blender animations with Opus, Astra and Argon: how you prompt still matters — chongdashu · 2026-10-16
- Running PolyStrata + Cline coding agent on a $0-fancy laptop at ~90 tok/s — rpwoerk · 2026-10-16
- How do you attribute agent gains: model training or infrastructure? — jrxna · 2026-10-16
- Feed Claude your books, blogs and top engineers' code for claude.md — 'it now writes better code than me' — blaizedsouza · 2026-10-16
- Open-source /label plugin auto-names and colors Claude Code sessions in one step — hermanschutte · 2026-10-16