INDEX

Explanations

social media actions related to sharing or communicating a message

oai_token-act-pair · gpt-3.5-turbo

instances of the word "posted" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

sth

-0.72

osi

-0.69

essen

-0.69

glers

-0.67

YD

-0.64

trak

-0.64

hya

-0.63

arge

-0.59

enser

-0.59

ichick

-0.58

POSITIVE LOGITS

 bulletin

0.88

 Instagram

0.80

edin

0.80

Redditor

0.78

 pics

0.76

 updates

0.75

information

0.75

 onto

0.73

 condol

0.72

 YouTube

0.72

Activations Density 0.183%

social media actions related to sharing or communicating a message

instances of the word "posted" in various contexts

No Comments

No Known Activations