2 papers
cs.LG2026
A Few Teacher Steps Go a Long Way: Cost-Efficient On-Policy Data Augmentation for Agent Post-Training
Junze Ye, Jiayi Cheng, Miao Lu +3
For LLM agents, supervised fine-tuning is not only about teacher labels' quality, but also about which interaction contexts those labels condition on. Pure behavioral cloning uses…
cs.AI2026
Scalable Stewardship of an LLM-Assisted Clinical Benchmark with Physician Oversight
Junze Ye, Daniel Tawfik, Alex J. Goodell +3
Reference labels for machine-learning benchmarks are increasingly synthesized with LLM assistance, but their reliability remains underexamined. We audit MedCalc-Bench, a clinical b…