INDEX

Explanations

instances of the word "aid" at varying activation levels

oai_token-act-pair · gpt-3.5-turbo

references to humanitarian aid

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

Negative Logits

 Bellev

-0.80

 Beard

-0.67

Ran

-0.66

theless

-0.65

 Mamm

-0.65

é¾

-0.64

 aber

-0.63

 archived

-0.61

 unforgettable

-0.61

 Wilde

-0.61

POSITIVE LOGITS

aid

1.37

glers

1.10

Aid

1.10

Aid

0.97

 aids

0.92

uese

0.89

maid

0.83

ãĥ¼ãĥĨ

0.81

aid

0.79

Reviewer

0.78

Activations Density 0.020%

instances of the word "aid" at varying activation levels

references to humanitarian aid

No Known Activations