AI.info

Tsui-Wei Weng

Explore Tsui-Wei Weng on AI.info.

• Data-Efficient Adaptation of LLMs via Attention Head Reweighting
• ReflCtrl: Controlling LLM Reflection Efficiently via Representation Engineering
• Resting Neurons, Active Insights: Robustifying Activation Sparsity in LLMs via Spontaneity
• Interpretable and Steerable Concept Bottleneck Sparse Autoencoders
• ReFIne: A Framework for Trustworthy Large Reasoning Models with Reliability, Faithfulness, and Interpretability