© Neuronpedia 2026

Privacy & Terms Blog GitHub Slack Twitter Contact

Neuronpedia

Jacobian LensNEW

Natural Language

NEW Assistant AxisNEW Circuit TracerUPDATESteer SAE Evals Exports interp-engineNEWAPI Community Blog Privacy & Terms Contact

GPT2-SMALL · 8-TRES-DC · 2136 ｜ Neuronpedia

Home
GPT2-Small
Transcoders Residuals
8-TRES-DC
2136

INDEX

Explanations

phrases indicating importance or value

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Embeds

Show PlotsShow ExplanationShow ActivationsShow Test FieldShow SteerShow Link

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

quer

-0.67

rez

-0.67

 LEGO

-0.64

MJ

-0.63

tz

-0.61

daq

-0.61

rul

-0.61

\-

-0.60

tub

-0.60

YD

-0.59

POSITIVE LOGITS

 utmost

0.88

 importance

0.86

onis

0.85

 relevance

0.85

 significance

0.80

 prominence

0.77

ificant

0.76

 paramount

0.71

 urgency

0.71

lev

0.70

Activations Density 0.016%

No Known Activations