INDEX

Explanations

phrases related to legal or criminal investigations

oai_token-act-pair · gpt-3.5-turbo

references to legal or investigative contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

ipel

-0.87

stall

-0.72

 helicop

-0.69

sonian

-0.68

perm

-0.68

 ãĤµ

-0.68

jin

-0.67

 preached

-0.65

oiler

-0.64

inen

-0.63

POSITIVE LOGITS

 thereto

1.03

 dealings

0.79

 dots

0.77

 proceedings

0.72

 laundering

0.71

 mater

0.69

llah

0.68

ities

0.68

ance

0.67

eering

0.66

Activations Density 0.023%

phrases related to legal or criminal investigations

references to legal or investigative contexts

No Comments

No Known Activations