Writing

• 12 of 15 Published Margins Are Smaller Than Their Own Benchmark Can See paper · reward-models · evaluation
• When Your Reward Model Cannot Matter: a $2 Measurement Before Training paper · reward-models · RL · scaling · evaluation
• GRPO Has a Silent Failure Mode. Here's How to Spot It. blog · RL · negative-results
• Same RL Recipe, Different Seed, Different Verdict: Here's a Training Comparison You Can Trust blog · reward-models · evaluation
• We Ranked Seven Reward Models. The Ranking Didn't Pick the Best Trainer. Here's the Check to Run Before You Choose Yours. blog · reward-models · RL
• Five Tools RL Training Libraries Should Ship blog · tooling

No posts with that tag yet.