Advanced Reinforcement Learning Interview Questions, issue 16, Feb 11, 2026

The Bootstrapping Bias Trap

Senior RL Engineer interview at OpenAI, and the interviewer asks:

We accidentally initialized our Value Network to output -1000 for every state. We run one update step using Monte Carlo and one using Bootstrapping (TD-Learning). Which algorithm breaks immediately, and which one survives?

Bootstrapping doesn’t just reduce variance, it injects your model’s current errors directly into the label, turning bad initialization into self-reinforcing policy collapse.

The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.

Read it on Substack

Get the next one

Free on Substack. Unsubscribe in one click.