Learning How Hard to Think: Input-Adaptive Allocation of LM Computation.

Mehul Damani, Idan Shenfeld, Andi Peng, Andreea Bobu, Jacob Andreas

Browse the full ICLR paper archive.