INDEX

Explanations

politically related terms, particularly names of politicians and government positions

oai_token-act-pair · gpt-3.5-turbo

commas in a variety of contexts within the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

reb

-0.75

ogle

-0.71

rup

-0.69

amas

-0.69

oric

-0.68

orts

-0.68

ãĥ¥

-0.68

orc

-0.66

lÃ©

-0.66

ibilities

-0.65

POSITIVE LOGITS

 namely

1.44

 albeit

1.14

viz

1.04

 however

0.91

 though

0.82

albeit

0.76

 according

0.73

 although

0.72

 alas

0.70

 perhaps

0.68

Activations Density 0.217%

politically related terms, particularly names of politicians and government positions

commas in a variety of contexts within the text

No Comments

No Known Activations