INDEX

Explanations

proper nouns related to individuals in various professional fields

oai_token-act-pair · gpt-3.5-turbo

references to former political figures

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

acht

-0.86

antics

-0.81

otle

-0.80

Ã±

-0.73

anche

-0.72

ammy

-0.72

atur

-0.72

atics

-0.71

mop

-0.67

ahon

-0.67

POSITIVE LOGITS

ly

0.95

 Yugoslav

0.92

 Former

0.79

 Vice

0.77

 Yugoslavia

0.76

 Deputy

0.74

bies

0.74

 former

0.71

 President

0.70

foundland

0.70

Activations Density 0.028%

proper nouns related to individuals in various professional fields

references to former political figures

No Comments

No Known Activations