Predicting Post-Training Coding-Agent Performance
Researchers propose methods for predicting whether base checkpoints are worth post-training for agentic coding tasks.
Researchers propose methods for predicting whether base checkpoints are worth post-training for agentic coding tasks.