INDEX

Explanations

promotional messages, social media platform names, and website URLs

oai_token-act-pair · gpt-3.5-turbo

mentions of social media platforms and interactions

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

 recounted

-0.71

 anesthesia

-0.66

 privatization

-0.64

EGIN

-0.64

 disputed

-0.63

cise

-0.62

 autonomy

-0.62

 encaps

-0.62

 sequencing

-0.62

 deregulation

-0.61

POSITIVE LOGITS

 facebook

1.57

 twitter

1.45

facebook

1.39

 Discord

1.38

 Patreon

1.34

Twitter

1.30

Facebook

1.28

 twitch

1.27

 Youtube

1.25

 Twitter

1.24

Activations Density 0.331%

promotional messages, social media platform names, and website URLs

mentions of social media platforms and interactions

No Comments

No Known Activations