© 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 · 1940 ｜ Neuronpedia

Home
GPT2-Small
Transcoders Residuals
8-TRES-DC
1940

INDEX

Explanations

repeated special characters or symbols that may indicate formatting or emphasis

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

ID

-0.75

 quote

-0.74

"[

-0.74

 verse

-0.73

 transcript

-0.71

-0.70

 minutes

-0.69

ut

-0.69

 year

-0.68

 root

-0.67

POSITIVE LOGITS

Magikarp

0.96

clinton

0.96

Offline

0.93

pots

0.92

EF

0.92

Pokemon

0.90

tackle

0.90

Mi

0.89

Arsenal

0.89

Va

0.89

Activations Density 0.088%

No Known Activations