Wonjun Kang

Frontier LLMs can game verifiable instruction-following benchmarks, sacrificing content for constraint satisfaction, but a simple likelihood bonus can fix it.

Asynchronous OPD can boost training throughput significantly while managing the challenges of stale data, transforming the efficiency of large language model fine-tuning.

EfficientRollout slashes RL rollout latency by nearly 20% while maintaining model performance, revolutionizing how we approach decoding in reinforcement learning.