INDEX

Explanations

mentions of geopolitical entities or territories

oai_token-act-pair · gpt-3.5-turbo

references to geographical regions or political areas

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

lder

-0.77

iris

-0.69

apego

-0.66

ppel

-0.66

CHA

-0.65

Bird

-0.65

ffer

-0.63

nea

-0.63

Oracle

-0.62

Enough

-0.62

POSITIVE LOGITS

hips

1.21

 territories

1.10

 Territories

0.99

 Territory

0.78

hops

0.78

enaries

0.78

afety

0.76

ystem

0.75

querque

0.73

awatts

0.73

Activations Density 0.006%

mentions of geopolitical entities or territories

references to geographical regions or political areas

No Comments

No Known Activations