INDEX

Explanations

mentions of individuals or groups related to news reporting or investigations

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "rep" and variations of it

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

 plausible

-0.59

 congr

-0.57

 deceptive

-0.56

 accur

-0.55

 explicit

-0.55

 daring

-0.55

 disturbing

-0.54

 tantal

-0.53

apt

-0.53

 generously

-0.53

POSITIVE LOGITS

icity

0.81

ications

0.78

lain

0.76

iates

0.71

igham

0.71

emort

0.70

swick

0.69

icator

0.67

ient

0.67

agues

0.67

Activations Density 0.157%

mentions of individuals or groups related to news reporting or investigations

instances of the prefix "rep" and variations of it

No Comments

No Known Activations