INDEX

Explanations

phrases related to continuing a story or newsletter signup

oai_token-act-pair · gpt-3.5-turbo

references to stories or articles

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

inence

-0.78

ateurs

-0.77

 stocking

-0.77

heed

-0.77

omin

-0.75

ometric

-0.75

yip

-0.75

utive

-0.74

alus

-0.74

ependence

-0.74

POSITIVE LOGITS

article

0.93

 Stories

0.81

 REPORT

0.81

story

0.80

Interstitial

0.80

Article

0.80

arc

0.80

fare

0.80

 ARTICLE

0.79

 Transcript

0.77

Activations Density 0.033%

phrases related to continuing a story or newsletter signup

references to stories or articles

No Comments

No Known Activations