INDEX

Explanations

mentions of political leaders, specifically the title of 'Speaker' and their associated names

oai_token-act-pair · gpt-3.5-turbo

references to political leaders and their roles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

eros

-0.79

schild

-0.73

ter

-0.71

arial

-0.67

ibaba

-0.66

asca

-0.66

kus

-0.65

tering

-0.64

erial

-0.64

agara

-0.64

POSITIVE LOGITS

 Speaker

1.00

 speaker

0.95

ħĭ

0.81

 Newt

0.75

loo

0.72

 laureate

0.72

 Gingrich

0.71

person

0.69

 Boehner

0.69

llor

0.68

Activations Density 0.008%

mentions of political leaders, specifically the title of 'Speaker' and their associated names

references to political leaders and their roles

No Comments

No Known Activations