Tag

#reward-hacking

Posts tagged with reward-hacking

2 posts

Verifiable rewards draw the reinforcement learning signal from checks already running in the stack, such as schemas, constraints, and test suites.

Engineering • Oct 23, 2026 •13 min read

Verifiable Rewards Are Already in Your Stack

Verifiable rewards are the real blocker in agent RL. Learn to mine database schemas, constraints, and tests for reward signals, plus when to train at all.

#reinforcement-learning #verifiable-rewards +4

The agent harness, with its routing, retries, and tool schemas around the model, defines the behavior that reinforcement learning actually optimizes.

Engineering • Oct 09, 2026 •11 min read

Agentic Reinforcement Learning in the Harness You Ship

Agentic reinforcement learning belongs in the harness you ship, not a training clone. Learn the five failure modes and when to skip RL entirely.

#agentic-rl #agent-lightning +4