INDEX

Explanations

words related to being user-friendly or favorable for use by a certain group

oai_token-act-pair · gpt-3.5-turbo

references to user-friendliness

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

ipt

-0.82

duct

-0.80

skill

-0.80

hid

-0.78

Bone

-0.78

hap

-0.75

ulet

-0.74

angan

-0.73

attr

-0.71

orah

-0.71

POSITIVE LOGITS

friendly

1.00

 Friendly

0.92

 friendly

0.87

glers

0.76

 confines

0.73

 darling

0.72

itably

0.69

yip

0.69

 Yankee

0.68

lier

0.68

Activations Density 0.010%

words related to being user-friendly or favorable for use by a certain group

references to user-friendliness

No Comments

No Known Activations