Fang Kong

Transformer context for in-context RL can be compressed to the Krylov dimension, revealing a $Θ(γ^m)$ decay in approximation error below this threshold.

Frontier LLMs can be trained to selectively ignore unreliable external guidance, a capability crucial for robust agentic behavior beyond simple task accuracy.

Offline data from outdated environments can still yield optimal policies when integrated correctly, as shown by our new hybrid reinforcement learning framework.

Teaching emotional support chatbots specific, executable skills, rather than relying on end-to-end training, leads to more interpretable, controllable, and ultimately more helpful conversations.

Allowing multiple support strategies in a single utterance can dramatically enhance the quality of emotional support conversations, leading to more effective dialogue outcomes.