INDEX

Explanations

names or terms related to political and military operatives

oai_token-act-pair · gpt-3.5-turbo

terms related to individuals involved in covert operations or espionage

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

xual

-0.73

yards

-0.70

Hig

-0.68

ranged

-0.67

este

-0.66

hus

-0.66

western

-0.66

etsk

-0.66

perm

-0.65

oran

-0.65

POSITIVE LOGITS

 operative

0.97

 posing

0.79

 operatives

0.78

doms

0.71

wear

0.70

 Emails

0.70

leck

0.67

 extraord

0.67

 outfits

0.67

 tasked

0.66

Activations Density 0.009%

names or terms related to political and military operatives

terms related to individuals involved in covert operations or espionage

No Comments

No Known Activations