CURIOUS ABOUT MACHINES. BUILDING SOME.

We audit the data robots learn from — and find what's broken in it.

A robot policy is only as good as the data underneath it. Almost nobody checks the data. We do — episode by episode, tracing each defect from the broken record back to the broken behavior.

broken data broken behavior follow the trace →
// PRACTICE

Trace Machines is independent robotics research. Right now the work is forensic: auditing the datasets that robot foundation models are trained on, and tracing the defects hidden inside them. It's where we're starting, not the edge of what this becomes.

How the audit works

// APPROACH
01

Read the data.

Go through episodes directly instead of trusting the summary stats. Assume the dataset is lying until checked.

02

Trace the defect.

Find where a flaw originates, which episodes it touches, and how it would corrupt training downstream.

03

Report it straight.

Neutral and buyer-side, with reproducible evidence. No incentive to flatter the data.

Work

// ARTIFACTS
// CONTACT

Working on a robot policy? Let's look at the data underneath it.