Wei Wu

Forget O(N^5) scaling: this new tensor decomposition approach slashes the computational cost of second-order perturbation theory down to O(N^3) without sacrificing accuracy.

LLM agents learn tool use far more efficiently when their learning environment evolves alongside them, adapting to the agent's changing capabilities.

Standard RLVR's token-level updates are often dominated by irrelevant tokens, but DelTA's discriminative reweighting unlocks significant performance gains in reasoning tasks.