INDEX

Explanations

Twitter handles to follow

oai_token-act-pair · gpt-3.5-turbo

references to social media, particularly Twitter

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

•twitter related features

•Twitter features

No Comments

Negative Logits

raints

-0.78

etheless

-0.65

 forgiven

-0.64

Cooldown

-0.62

bably

-0.62

forced

-0.60

 Meaning

-0.59

venge

-0.58

 antidepressants

-0.57

stood

-0.57

POSITIVE LOGITS

(@

0.96

0.91

 twitter

0.88

edin

0.86

 âĿ

0.84

 hasht

0.78

 Twitter

0.78

Tweet

0.76

 Tweet

0.75

facebook

0.75

Activations Density 0.044%

Twitter handles to follow

references to social media, particularly Twitter

No Comments

No Known Activations