INDEX

Explanations

phrases containing the word "onde"

oai_token-act-pair · gpt-3.5-turbo

references to geographic locations or landmarks

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

shr

-0.73

Ï

-0.69

 scissors

-0.67

div

-0.66

 Squirrel

-0.65

 Invaders

-0.64

 Magic

-0.64

jew

-0.63

 Legal

-0.61

 snakes

-0.59

POSITIVE LOGITS

onde

4.98

ond

1.76

ondo

1.37

ande

1.28

onda

1.25

onds

1.23

arde

1.17

alde

1.16

OND

1.13

undo

1.09

Activations Density 0.015%

phrases containing the word "onde"

references to geographic locations or landmarks

No Comments

No Known Activations