Getting More Juice Out of the SFT Data: Reward Learning from Human Demonstration Improves SFT

Best AI papers explained - Getting More Juice Out of the SFT Data: Reward Learning from Human Demonstration Improves SFT