INDEX

Explanations

the word "Bio"

oai_token-act-pair · gpt-3.5-turbo

references to bio-related concepts or terms

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

lean

-0.78

 Relax

-0.76

zy

-0.73

Zed

-0.66

 Yates

-0.65

 Squirrel

-0.64

igans

-0.63

 McCl

-0.63

ishly

-0.63

RAM

-0.63

POSITIVE LOGITS

Bio

1.18

 circum

0.73

 worth

0.71

ographed

0.70

lif

0.68

Detroit

0.68

iph

0.67

orbit

0.66

piece

0.65

accompan

0.63

Activations Density 0.001%

the word "Bio"

references to bio-related concepts or terms

No Comments

No Known Activations