INDEX

Explanations

words and phrases related to LGBTQ+ identities, specifically focusing on bisexuality

oai_token-act-pair · gpt-3.5-turbo

references to bisexuality

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

EMS

-0.82

urers

-0.81

 Amend

-0.71

LV

-0.71

Eat

-0.71

 Dept

-0.68

earchers

-0.67

witz

-0.67

OOL

-0.67

×ķ

-0.67

POSITIVE LOGITS

ity

1.09

isexual

0.84

iliary

0.82

inals

0.79

icide

0.79

 bisexual

0.77

iliated

0.76

dar

0.75

 citiz

0.71

ikini

0.69

Activations Density 0.010%

words and phrases related to LGBTQ+ identities, specifically focusing on bisexuality

references to bisexuality

No Comments

No Known Activations