INDEX

Explanations

phrases related to news and alerts

oai_token-act-pair · gpt-3.5-turbo

phrases related to breaking news announcements

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

inating

-0.75

yon

-0.74

itary

-0.72

ivan

-0.71

haps

-0.71

rals

-0.70

ung

-0.69

wagen

-0.68

ript

-0.68

itance

-0.68

POSITIVE LOGITS

fast

1.07

AKING

0.83

 NEWS

0.82

 news

0.82

Bad

0.82

 News

0.79

Berry

0.78

NEWS

0.74

views

0.73

 Breaking

0.72

Activations Density 0.026%

phrases related to news and alerts

phrases related to breaking news announcements

No Comments

No Known Activations