Skip to content

LLM Transparency Tool

An open-source toolkit for looking inside transformer models

Visit official site

What LLM Transparency Tool is

The toolkit is described as an open-source interactive tool for analysing the internal workings of transformer language models, aimed at improving model transparency.

What you can do with it

  • Pick a model and a prompt to run inference
  • Browse contribution graphs and adjust the threshold
  • Build a graph from selected tokens

Model notes

  • Representations are available after any block
  • Output vocabulary representations are shown
  • Promoted and suppressed tokens from the previous block are listed

Pros & cons

✓ What we like

  • Representations can be selected after any block
  • Output vocabulary representations show promoted or suppressed tokens

! What to watch out for

  • The material lists no setup steps for this entry
  • Analysis starts from a chosen model and prompt

FAQ

What does it visualise?

The material says it shows attention flow and information transfer paths.

Can tokens be inspected?

The material lists building graphs from tokens and viewing representations after any block.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review