INDEX

Explanations

words related to specific topics or categories, potentially related to health, technology, and community

oai_token-act-pair · gpt-3.5-turbo

the word "For" with varying contexts indicating lists or introductions

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

è¦ļéĨĴ

-0.71

Zen

-0.63

æĺ¯

-0.60

soDeliveryDate

-0.60

illin

-0.60

inese

-0.59

>"

-0.58

EStreamFrame

-0.58

$.

-0.57

Introduced

-0.57

POSITIVE LOGITS

gotten

1.37

bidden

1.32

cing

1.28

 example

1.22

ced

1.14

 instance

1.11

WARD

1.07

gery

1.05

give

1.00

wards

0.98

Activations Density 0.105%

words related to specific topics or categories, potentially related to health, technology, and community

the word "For" with varying contexts indicating lists or introductions

No Comments

No Known Activations