INDEX

Explanations

phrases related to initiating or beginning something

oai_token-act-pair · gpt-3.5-turbo

references to athletes or performers in a starting position

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

asca

-0.74

yss

-0.70

erry

-0.70

orate

-0.68

oÄŁ

-0.68

fax

-0.67

iq

-0.65

erm

-0.65

umb

-0.65

 controlled

-0.64

POSITIVE LOGITS

 starters

1.09

starter

0.87

starting

0.86

 starter

0.85

 pitchers

0.81

 thereof

0.80

éļ

0.75

 Clicker

0.73

vation

0.72

 isEnabled

0.70

Activations Density 0.002%

phrases related to initiating or beginning something

references to athletes or performers in a starting position

No Comments

No Known Activations