INDEX

Explanations

phrases related to news headlines and topics like continued articles, elections, and human rights

oai_token-act-pair · gpt-3.5-turbo

references to articles or sections within articles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Vaugh

-0.28

 Azerb

-0.25

rul

-0.24

 disadvant

-0.22

 challeng

-0.22

vertisement

-0.20

 Constantin

-0.20

 destro

-0.20

'."

-0.20

 Ambro

-0.19

POSITIVE LOGITS

 âĢº

0.25

conom

0.24

utterstock

0.22

pmwiki

0.20

taboola

0.20

Baltimore

0.20

cigarettes

0.20

Sab

0.20

natureconservancy

0.19

ratom

0.19

Activations Density 7.729%

phrases related to news headlines and topics like continued articles, elections, and human rights

references to articles or sections within articles

No Comments

No Known Activations