Instrument

R. Luger; Harish Kamath; Doug Finkbeiner; Purvi Goel; Adam Jermyn; Sam Zimmerman; Joshua Batson; Tom Conerly · Anthropic

An interactive tool for interrogating attention heads, released as the research output in its own right. Its central finding is methodological: a head's behaviour on the full data distribution rarely matches what a narrow task suggests.

Reframes the tool as the contribution, and supplies the concrete examples the field needs to attack attention decomposition.

High · Read the full article, 2026-08-06

Yida Chen; Aoyu Wu; Trevor DePodesta; Catherine Yeh; Kenneth Li; Nicholas Castillo Marin; Oam Patel; Jan Riecke; Shivam Raval; Olivia Seow; Martin Wattenberg; Fernanda Viégas · Harvard University

A dashboard that sits beside a chatbot and shows, in real time, what the model has internally inferred about the user's age, gender, education and socioeconomic status — and lets the user edit those inferences and watch the responses change.

Almost everything else in this ledger is built for researchers; this is built for the person being modelled, and a user study found it helped participants expose the system's biased behaviour.

High · Read the arXiv abstract page in full, 2026-08-06. Project page (bit.ly/talktuner-project-page) not fetched.

Catherine Yeh; Yida Chen; Aoyu Wu; Cynthia Chen; Fernanda Viégas; Martin Wattenberg · Harvard University

Visualises a joint embedding of the query and key vectors a transformer uses to compute attention, which makes it possible to see global patterns across many input sequences rather than one prompt at a time.

Shifted attention visualisation from single-example to distribution-level — the exact move Anthropic's HeadVis names as its closest prior work three years later.

High · Read the arXiv abstract page in full, 2026-08-06. The live demo at attentionviz.com was not inspected — client-rendered, fetch returned an empty page shell.

Johnny Lin (creator), with community contributors · Neuronpedia / Decode Research

The field's central public platform. It hosts feature dashboards, attribution graphs, steering and demos across dozens of open models — including the official interactive releases for Anthropic's Jacobian Lens, Natural Language Autoencoders, Assistant Axis and Circuit Tracer, and DeepMind's Gemma Scope.

The closest thing interpretability has to a public commons, and the reason a frontier-lab result can now be poked at by an outsider the week it ships.

High · Read the neuronpedia.org homepage in full, 2026-08-06