INDEX

Explanations

words related to strangeness or peculiarity

oai_token-act-pair · gpt-3.5-turbo

instances of the word "weird."

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

 commenced

-0.87

nor

-0.83

 Administration

-0.75

abil

-0.70

ILE

-0.70

due

-0.70

 plaintiff

-0.68

PLA

-0.68

 âĢİ

-0.67

 sufficient

-0.67

POSITIVE LOGITS

 weird

3.00

 Weird

2.16

 creepy

1.64

 shitty

1.61

 craz

1.60

 strang

1.56

 goofy

1.55

 bizarre

1.53

odd

1.52

 awkward

1.52

Activations Density 0.035%

words related to strangeness or peculiarity

instances of the word "weird."

No Comments

No Known Activations