INDEX

Explanations

the word "ottawa"

oai_token-act-pair · gpt-3.5-turbo

repeated patterns of the same substring, particularly words that contain "otta" and "arella."

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

termin

-0.73

Hop

-0.66

eways

-0.64

sl

-0.63

Department

-0.63

circ

-0.63

Closure

-0.62

drawn

-0.62

travel

-0.62

Adv

-0.61

POSITIVE LOGITS

otta

1.14

onna

0.91

chio

0.89

hedral

0.88

acus

0.86

asma

0.85

DonaldTrump

0.85

hler

0.84

uria

0.82

olini

0.81

Activations Density 0.006%

the word "ottawa"

repeated patterns of the same substring, particularly words that contain "otta" and "arella."

No Comments

No Known Activations