INDEX

Explanations

names such as "Brian Hagen" or variations of "agen" with high activation values

oai_token-act-pair · gpt-3.5-turbo

references to aging and age-related concepts

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

 Kinnikuman

-0.67

 Reese

-0.66

isSpecialOrderable

-0.62

 Doct

-0.60

 Heard

-0.60

erest

-0.59

 Investigator

-0.58

 Gamble

-0.58

HIP

-0.57

 Season

-0.57

POSITIVE LOGITS

furt

1.44

esis

1.31

agen

1.08

etic

1.02

etics

0.97

arian

0.92

hess

0.89

arians

0.89

omics

0.88

alogy

0.83

Activations Density 0.007%

names such as "Brian Hagen" or variations of "agen" with high activation values

references to aging and age-related concepts

No Comments

No Known Activations