INDEX

Explanations

phrases related to negative or derogatory descriptions

oai_token-act-pair · gpt-3.5-turbo

references to holes, particularly in a metaphorical or humorous context

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ency

-0.83

ERG

-0.78

ables

-0.70

ione

-0.69

lux

-0.69

Joy

-0.68

ICT

-0.67

REE

-0.66

 Carbuncle

-0.66

RON

-0.65

POSITIVE LOGITS

hole

1.56

holes

1.29

 Hole

0.94

 hole

0.91

ocene

0.88

shit

0.84

 holes

0.81

hog

0.79

izons

0.75

 Cerberus

0.73

Activations Density 0.014%

phrases related to negative or derogatory descriptions

references to holes, particularly in a metaphorical or humorous context

No Comments

No Known Activations