INDEX

Explanations

keywords related to technology and product names

oai_token-act-pair · gpt-3.5-turbo

references to "zeros" and "PD" (possibly referring to professional designations or specific categories)

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

shadow

-0.69

activity

-0.66

Pv

-0.64

âĦ¢:

-0.64

rail

-0.63

 abduction

-0.60

classes

-0.59

 Ward

-0.57

 sculpt

-0.56

Fe

-0.56

POSITIVE LOGITS

eros

4.68

ERO

1.64

eno

0.89

cest

0.89

erie

0.87

gins

0.86

verages

0.82

enos

0.81

er

0.80

yip

0.80

Activations Density 0.019%

keywords related to technology and product names

references to "zeros" and "PD" (possibly referring to professional designations or specific categories)

No Comments

No Known Activations