Elemental Research

Understand AI
from the inside.

We study the internal logic of intelligent systems: what they represent, how they predict, and why they act. Our work turns those questions into tools and testable research.

Research Projects

Projects

OPEN SOURCE TOOLWorldModelLensObserve internal states and replay model trajectories. RESEARCH FRAMEWORKAttribution–Attention FaithfulnessTest whether highlighted signals actually shape predictions. IN DEVELOPMENTEval SuiteBuild reproducible comparisons of model behavior.