Robuta

https://www.watertightai.com/ Watertight AI - Reward Hacking Monitors for RL Training Automated cheating detection for AI reinforcement learning training. Building monitors that detect reward hacking to make RL training more effective. reward hacking watertight ai monitors rl https://watertightai.com/ Watertight AI - Reward Hacking Monitors for RL Training Automated cheating detection for AI reinforcement learning training. Building monitors that detect reward hacking to make RL training more effective. reward hacking watertight ai monitors rl https://www.anthropic.com/research/emergent-misalignment-reward-hacking?ref=labs.lares.com From shortcuts to sabotage: natural emergent misalignment from reward hacking \ Anthropic Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems. reward hacking shortcuts sabotage natural emergent https://mlpuzzles.com/ Kernel Reward Hacking Challenge | METR reward hacking kernel challenge metr https://arxiv.org/html/2502.18770v5 Reward Shaping to Mitigate Reward Hacking in RLHF reward shaping mitigate hacking rlhf