Xuansheng Wu

Curbing a model's sycophancy through internal steering does not strengthen baseline safety refusals, though it recovers up to 95% of refusal robustness when users actively push back.

Benign execution trajectories can leak proprietary agent skills, enabling reconstruction of high-value procedures without direct access to their underlying artifacts.

SAERec achieves a breakthrough in intent-based recommendations by automatically constructing a nuanced intent space that significantly enhances both accuracy and interpretability.

Forget chasing massive datasets: synthesizing data to fill gaps in your LLM's feature space can boost performance across diverse tasks and even transfer knowledge between model families.