Chenhao Dang

2 accepted papers

2026

AC-ODM: Actor–Critic Online Data Mixing for Sample-Efficient LLM Pretraining

ICML 2026poster

Optimizing pretraining data composition is pivotal for LLM generalization. While dynamic mixing outperforms static strategies by capturing evolving training dynamics, current methods fail to reconcile computational efficiency with sample efficiency and structural flexibility for diverse pipelines.We…

Cited by 0 Source ↗Scholar ↗

2026

Breaking the Adversarial Robustness-Performance Trade-off in Text Classification via Manifold Purification

Chenhao Dang, Jing Ma

AAAI 2026technical

A persistent challenge in text classification (TC) is that enhancing model robustness against adversarial attacks typically degrades performance on clean data. We argue that this challenge can be resolved by modeling the distribution of clean samples in the encoder’s embedding manifold. To this end,

Cited by 0 Source ↗PDF ↗Scholar ↗