INDEX

Explanations

dates and names mentioned in the text

oai_token-act-pair · gpt-3.5-turbo

references to dates and authors in a citation format

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 congr

-0.70

 Creed

-0.69

 vows

-0.64

 Angle

-0.64

 Dame

-0.64

 Hail

-0.63

igans

-0.63

aceae

-0.63

 refunds

-0.63

Ap

-0.62

POSITIVE LOGITS

BBC

1.48

WASHINGTON

1.07

Associated

1.03

CBS

0.98

NEW

0.93

BER

0.92

remlin

0.91

PRESS

0.90

Reuters

0.90

CNN

0.89

Activations Density 0.059%

dates and names mentioned in the text

references to dates and authors in a citation format

No Comments

No Known Activations