Skip to content

Transformer Debugger (TDB)

Investigate small language model behaviour

Visit official site

What Transformer Debugger (TDB) is

The tool is described as an OpenAI instrument for investigating specific behaviours of small language models.

What you can do with it

  • Identify components behind a behaviour
  • Observe how an intervention changes a behaviour

Pros & cons

✓ What we like

  • Automated explanations are generated
  • Neurons, attention heads and autoencoder latents are named

! What to watch out for

  • It targets small language models
  • Investigation relies on interpretability components

FAQ

Who developed it?

The material names the OpenAI Superalignment team.

What does it explain?

It shows why specific components activate strongly.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review