INDEX

Explanations

terms related to unhealthy or low-quality items such as "junk food" and "junk bonds"

oai_token-act-pair · gpt-3.5-turbo

references to "junk" in various contexts

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

Æ

-0.89

cing

-0.68

ndum

-0.64

 Scarlett

-0.64

BILITY

-0.63

 Kubrick

-0.63

 Wilkinson

-0.60

 Rhythm

-0.59

 Tsarnaev

-0.59

sis

-0.59

POSITIVE LOGITS

 junk

1.19

abies

0.87

rats

0.83

 dumps

0.82

adelphia

0.80

busters

0.78

buster

0.78

vertisement

0.78

-+-+

0.76

 drawer

0.76

Activations Density 0.007%

terms related to unhealthy or low-quality items such as "junk food" and "junk bonds"

references to "junk" in various contexts

No Comments

No Known Activations