# facebookresearch/llm-transparency-tool

LLM Transparency Tool (LLM-TT), an open-source interactive toolkit for analyzing internal workings of Transformer-based language models. *Check out demo at* https://huggingface.co/spaces/facebook/llm-transparency-tool-demo

Repository: https://github.com/facebookresearch/llm-transparency-tool
Canonical: https://ross.abutalabs.com/products/llm-transparency-tool
Language: Python
License: NOASSERTION
License Family: other
Archived: true
Last push: 2024-12-03T22:41:46+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 70
- inputs: {"age_days": 986, "days_push": 638, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1254, forks 107 (observed 2026-08-28T04:04:08.691054+00:00)

## What it is
An interactive toolkit from Meta Research for analyzing the internal workings of Transformer-based language models. It visualizes contribution graphs of tokens, attention heads, and FFN neurons through a Streamlit web UI to explain model predictions.

## Use cases
- inspect how a transformer LLM processes a prompt internally
- visualize token contribution graphs for model interpretability
- see which attention heads promote or suppress output tokens
- debug why an LLM made a certain prediction
- analyze FFN neuron activations in a language model
- research tool for mechanistic interpretability of transformers

## When to choose
- you need interactive, visual interpretability analysis of a TransformerLens-supported model
- you're doing research on how LLMs make predictions
- you want to inspect attention heads and FFN contributions per token

## When to avoid
- you need production LLM serving or inference at scale
- your model isn't supported by TransformerLens and you can't implement a custom TransparentLlm wrapper
- you need automated interpretability pipelines rather than interactive exploration

## Facets
- artifact type: application
- maturity: active
- function: machine-learning, llm-inference, data-visualization, developer-tools
- domain: large-language-models, machine-learning, deep-learning, data-visualization
- platform: python, cross-platform
- tags: interpretability, transformer-lens, streamlit, attention-analysis, contribution-graph, explainability, research-tool, research, docker, web-server

## Member repositories
- facebookresearch/llm-transparency-tool (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:08.691054+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T05:07:23.961383+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/llm-transparency-tool (fetched 2026-08-28T04:04:08.691054+00:00, sha 764c71408b3e)
- Data as of 2026-08-30T08:39:29.467469+00:00.