© Neuronpedia 2026

Privacy & Terms Blog GitHub Slack Twitter Contact

Neuronpedia

Jacobian LensNEW

Natural Language

NEW Assistant AxisNEW Circuit TracerUPDATESteer SAE Evals ExportsAPI Community Blog Privacy & Terms Contact

GPT2-SMALL · 8-TRES-DC · 540 ｜ Neuronpedia

Home
GPT2-Small
Transcoders Residuals
8-TRES-DC
540

INDEX

Explanations

phrases indicating potential advancements in treatments or technology

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

terness

-0.75

strate

-0.75

uggest

-0.69

覚醒

-0.69

セ

-0.65

andum

-0.65

auri

-0.64

wic

-0.64

abor

-0.64

 Calls

-0.64

POSITIVE LOGITS

eem

0.87

guard

0.69

iful

0.68

lap

0.68

Pred

0.67

eday

0.66

Blade

0.65

ghai

0.65

 Seraph

0.64

 proper

0.64

Activations Density 0.412%

No Known Activations