© 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 · 2500 ｜ Neuronpedia

Home
GPT2-Small
Transcoders Residuals
8-TRES-DC
2500

INDEX

Explanations

words related to poetry

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

lain

-0.68

lass

-0.67

rav

-0.66

nance

-0.65

uras

-0.65

unning

-0.64

rap

-0.63

ridge

-0.63

 rapt

-0.62

 assuming

-0.60

POSITIVE LOGITS

pp

0.82

pps

0.81

ppers

0.74

pees

0.73

pload

0.73

�

0.72

ciation

0.71

inters

0.71

lem

0.70

existence

0.69

Activations Density 0.011%

No Known Activations