20 papers
Which Decisions Low-Bit Quantization Breaks, and How to Predict Them
Zekun Wu, Swati Dhiman, Adriano Koshiyama
Quantization is known to hurt below four bits, but nobody can say which of a model's decisions will change at a given bit-width. This matters most where a model acts rather than an…
Routing Is Least Learnable Where It Is Most Valuable: Bounds on Representation Routing for Web Agents
Jiaming Wei, Zekun Wu, Adriano Koshiyama +1
Web agents observe a browser through text, pixels, or both, and the choice is usually fixed once for all tasks. We measure six observation modes across eight site-model combination…
Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects
Seonglae Cho, Zekun Wu, Kleyton Da Costa +3
Sparse autoencoder (SAE) features are used to interpret and steer large language models, yet whether a feature's causal role is stable across SAE families remains untested. Single-…
Tool Calling is Linearly Readable and Steerable in Language Models
Zekun Wu, Ze Wang, Seonglae Cho +4
When a tool-calling agent picks the wrong tool, the failure is invisible until execution: the email gets sent, the meeting gets missed. As agents take on consequential actions, one…
Sell Me This Stock: Unsafe Recommendation Drift in LLM Agents
Zekun Wu, Adriano Koshiyama, Sahan Bulathwela +1
People increasingly use LLM agents for multi-turn financial recommendations, where the agent pulls market data through tools and tracks user preferences across turns. When tool out…
Control Reinforcement Learning: Interpretable Token-Level Steering of LLMs via Sparse Autoencoder Features
Seonglae Cho, Zekun Wu, Adriano Koshiyama
Sparse autoencoders (SAEs) decompose language model activations into interpretable features, but existing methods reveal only which features activate, not which change model output…