INDEX

Explanations

mentions of a specific political party

oai_token-act-pair · gpt-3.5-turbo

references to the Liberal political party

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

catentry

-0.78

maps

-0.78

 torches

-0.73

crow

-0.73

 thru

-0.71

owl

-0.68

ember

-0.66

tyard

-0.66

raph

-0.65

 simulate

-0.65

POSITIVE LOGITS

 Liberal

3.73

 Liberals

3.08

Liberal

2.93

 Conservative

2.41

 Conservatives

2.34

NDP

2.10

Conservative

1.90

 Tories

1.81

 Tory

1.79

 Libertarian

1.69

Activations Density 0.014%

mentions of a specific political party

references to the Liberal political party

No Comments

No Known Activations