Enci Zhang / 肥鲶鱼

Studying how models learn.

I work on language model post-training, evaluation, and agent systems. This is where I share research, engineering work, and ideas in progress.

Selected Work

EMNLP 2025

Adaptive difficulty curricula and expert-guided self-reformulation for reasoning models.

NeurIPS 2025 Workshop

A study of stable, efficient validation signals for vision-language model training.

PRICAI 2025

Joint optimization of models and agent workflows under performance, cost, and efficiency objectives.

2026-09-15

A home for research, code, and ideas.