INDEX

Explanations

the word "Nice"

oai_token-act-pair · gpt-3.5-turbo

instances of the words "Nice" and "Grab."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 timely

-0.78

ocene

-0.78

 Appalachian

-0.69

onics

-0.65

 opposing

-0.63

ript

-0.63

psi

-0.62

Jem

-0.61

andestine

-0.61

itudinal

-0.60

POSITIVE LOGITS

 Nice

2.35

 Grab

1.89

Grab

1.64

Nice

1.57

grab

1.45

flow

1.37

 Braun

1.18

 Cannes

1.18

 Bild

1.16

Lug

1.13

Activations Density 0.052%

the word "Nice"

instances of the words "Nice" and "Grab."

No Comments

No Known Activations