AIAnthropicabout 5 hours ago
Anthropic Tests GRAFT to Stop Reward Hacking
6 min read
Anthropic researchers propose GRAFT, a method that localizes bad behavior in AI models by routing gradients to removable parameters. Testing on Qwen3-8B, the technique drops reward hacking rates to nearly zero while keeping correct solutions 20 percentage points higher than the base model, offering a safer alternative to penalizing detected hacks.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
