INDEX

Explanations

phrases or occurrences related to Afghanistan

oai_token-act-pair · gpt-3.5-turbo

mentions of particular geographical or national identities

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

 Emer

-0.81

 Grant

-0.79

 Lear

-0.79

Gay

-0.78

 Simon

-0.76

 Turnbull

-0.76

 Trudeau

-0.75

 Brand

-0.75

 Mason

-0.75

 Megan

-0.75

POSITIVE LOGITS

Cu

2.15

Af

2.14

Ca

2.03

Gi

1.84

Ta

1.83

Fe

1.79

Da

1.79

Ap

1.78

Ly

1.77

Az

1.77

Activations Density 0.039%

phrases or occurrences related to Afghanistan

mentions of particular geographical or national identities

No Comments

No Known Activations