INDEX

Explanations

words related to specific geographical locations, particularly those ending in 'row'

oai_token-act-pair · gpt-3.5-turbo

references to specific locations or geographical names

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 metic

-0.89

 nomine

-0.82

£ı

-0.82

 mosqu

-0.74

 Gujar

-0.73

 Nigerian

-0.71

 incent

-0.70

ãĥ¤

-0.69

 scapego

-0.65

 blown

-0.64

POSITIVE LOGITS

icz

1.18

restling

1.13

orld

1.10

ski

1.08

zee

0.99

dy

0.93

een

0.92

row

0.91

atts

0.88

shaw

0.83

Activations Density 0.016%

words related to specific geographical locations, particularly those ending in 'row'

references to specific locations or geographical names

No Comments

No Known Activations