INDEX

Explanations

politically related terms, particularly names and positions in government

oai_token-act-pair · gpt-3.5-turbo

references to political leadership roles and party affiliation in the government

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

alach

-0.82

axis

-0.79

icular

-0.75

imov

-0.67

iru

-0.66

ipolar

-0.65

yip

-0.64

autical

-0.64

jiang

-0.63

anco

-0.63

POSITIVE LOGITS

 Caucus

1.01

 Whip

0.83

 Leader

0.82

Senate

0.81

 Committee

0.80

 Majority

0.77

 Subcommittee

0.77

 Assembly

0.76

 Chair

0.76

 Minority

0.76

Activations Density 0.065%

politically related terms, particularly names and positions in government

references to political leadership roles and party affiliation in the government

No Comments

No Known Activations