INDEX

Explanations

the word "Hi" at different levels of emphasis

oai_token-act-pair · gpt-3.5-turbo

greetings and informal salutations

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

lust

-0.77

 compuls

-0.66

 destabil

-0.64

 deliber

-0.64

 matters

-0.64

bleacher

-0.63

cons

-0.62

 reconcil

-0.62

 carn

-0.62

 plotting

-0.62

POSITIVE LOGITS

Hi

3.81

Hi

3.11

hi

2.31

Hello

2.08

 Hello

1.73

Dear

1.46

hi

1.42

HI

1.41

hello

1.39

 hello

1.38

Activations Density 0.009%

the word "Hi" at different levels of emphasis

greetings and informal salutations

No Comments

No Known Activations