INDEX

Explanations

mentions of former individuals

oai_token-act-pair · gpt-3.5-turbo

references to individuals with a prior status or position

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

utra

-0.68

fw

-0.68

angles

-0.67

zai

-0.64

aph

-0.63

acs

-0.61

Fram

-0.61

TT

-0.61

grim

-0.61

ggle

-0.61

POSITIVE LOGITS

 former

3.32

 Former

2.29

Former

2.25

former

2.03

 longtime

1.71

 formerly

1.62

 retired

1.59

 latter

1.56

 veteran

1.42

 disgr

1.36

Activations Density 0.039%

mentions of former individuals

references to individuals with a prior status or position

No Comments

No Known Activations