INDEX

Explanations

the word "hidden" occurring with emphasis

oai_token-act-pair · gpt-3.5-turbo

terms related to concealment or secrecy

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

nia

-0.83

SHIP

-0.81

xual

-0.81

annis

-0.79

thood

-0.78

Pwr

-0.76

udeb

-0.74

urses

-0.74

sburgh

-0.73

ctive

-0.71

POSITIVE LOGITS

 hidden

1.01

 Hidden

0.87

 behind

0.87

 recess

0.85

 secret

0.83

 camoufl

0.83

 secrets

0.80

gem

0.79

 gems

0.79

 hatch

0.78

Activations Density 0.015%

the word "hidden" occurring with emphasis

terms related to concealment or secrecy

No Comments

No Known Activations