Can You Trust an LLM Judge? An RL Researcher's Take

Name: Can You Trust an LLM Judge? An RL Researcher's Take
Uploaded: 2026-03-10T04:00:53+00:00
Channel: Deep Learning with Yacine
Description: Zichen Liu from Dr. GRPO breaks down LLM-as-a-judge from an RL perspective: why it's essentially a model-based reward function, how it compares to veri...

Deep Learning with Yacine · Advanced ·🧠 Large Language Models ·2mo ago

Zichen Liu from Dr. GRPO breaks down LLM-as-a-judge from an RL perspective: why it's essentially a model-based reward function, how it compares to verification-based rewards, and why it can unlock dense rewards for reasoning tasks that rules simply can't verify. yacine is still suspicious.

Watch on YouTube ↗ (saves to browser)