INDEX

Explanations

the presence of the word "cover" or its variations related to concealing or hiding something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "cover" and its variations, indicating a focus on concealment or protection

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

----------

-0.66

bour

-0.64

EO

-0.64

friend

-0.61

ue

-0.61

********************************

-0.60

 CARD

-0.60

nir

-0.59

--------------------------------------------------------

-0.58

================================================================

-0.58

POSITIVE LOGITS

ages

0.83

up

0.79

alls

0.76

 bases

0.74

iday

0.71

usky

0.71

 extensively

0.69

 basics

0.68

 cover

0.68

 expense

0.66

Activations Density 0.049%

the presence of the word "cover" or its variations related to concealing or hiding something

instances of the word "cover" and its variations, indicating a focus on concealment or protection

No Comments

No Known Activations