INDEX

Explanations

words related to geographic locations or addresses

oai_token-act-pair · gpt-3.5-turbo

references to various locations, particularly those ending in "erton."

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Kh

-0.69

Wik

-0.65

doc

-0.65

 Vers

-0.62

 status

-0.61

 word

-0.61

sub

-0.60

ver

-0.60

ex

-0.59

DE

-0.58

POSITIVE LOGITS

erton

4.77

erville

1.83

enegger

1.49

arton

1.46

ert

1.18

atown

1.17

leton

1.17

enhagen

1.14

eston

1.14

hurst

1.11

Activations Density 0.017%

words related to geographic locations or addresses

references to various locations, particularly those ending in "erton."

No Comments

No Known Activations