INDEX

Explanations

the word "named" with various activations ranging from 8 to 10

oai_token-act-pair · gpt-3.5-turbo

words that denote being named or labeled

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

 assurance

-0.67

ŃĶ

-0.64

 probable

-0.63

 forge

-0.60

 buckle

-0.59

 lock

-0.59

 weave

-0.58

 grade

-0.58

vac

-0.57

 guarantee

-0.57

POSITIVE LOGITS

amed

1.51

ame

1.23

ames

1.17

aming

1.06

urally

0.98

amia

0.98

milo

0.91

iliary

0.88

amer

0.83

amy

0.83

Activations Density 0.005%

the word "named" with various activations ranging from 8 to 10

words that denote being named or labeled

No Comments

No Known Activations