INDEX

Explanations

experiences or situations described as being observed or encountered firsthand

oai_token-act-pair · gpt-3.5-turbo

references to personal experiences or firsthand accounts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ramid

-0.77

nam

-0.77

nan

-0.76

rams

-0.73

ishops

-0.72

dar

-0.71

separ

-0.69

nah

-0.69

hate

-0.68

corn

-0.67

POSITIVE LOGITS

 firsthand

1.35

 attest

0.82

ewitness

0.78

 eyewitness

0.76

ãĥ¥

0.71

 unden

0.71

é¾įåĸļå£«

0.70

 acquainted

0.68

 perspect

0.67

Effects

0.66

Activations Density 0.003%

experiences or situations described as being observed or encountered firsthand

references to personal experiences or firsthand accounts

No Comments

No Known Activations