INDEX

Explanations

phrases related to political positions or titles

oai_token-act-pair · gpt-3.5-turbo

negative political or military sentiment

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Dickinson

-0.97

ulhu

-0.87

 Lovecraft

-0.83

 Scrib

-0.79

 Burr

-0.78

 Savage

-0.78

 Vaughn

-0.77

DeV

-0.76

 Yelp

-0.75

 Melvin

-0.74

POSITIVE LOGITS

based

1.27

government

1.18

sized

1.09

backed

1.08

linked

1.07

induced

1.04

national

1.03

organ

1.02

equipped

1.02

mediated

1.01

Activations Density 0.345%

phrases related to political positions or titles

negative political or military sentiment

No Comments

No Known Activations