INDEX

Explanations

comments or messages written by users expressing their opinions

oai_token-act-pair · gpt-3.5-turbo

phrases regarding user comments and participation on social media

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

 headquartered

-0.74

 subord

-0.67

hart

-0.66

sac

-0.66

 enshr

-0.65

 deregulation

-0.63

 treaties

-0.63

 negotiating

-0.63

 Spending

-0.63

ÂłÂłÂłÂłÂłÂłÂłÂł

-0.63

POSITIVE LOGITS

Redditor

1.10

 commenters

1.07

 Reddit

1.04

Reddit

0.99

username

0.98

 feedback

0.96

ickr

0.93

estamp

0.85

 commenter

0.85

 submissions

0.83

Activations Density 0.656%

comments or messages written by users expressing their opinions

phrases regarding user comments and participation on social media

No Comments

No Known Activations