INDEX

Explanations

the word "ÅŁ" with various levels of activation strength

oai_token-act-pair · gpt-3.5-turbo

specific characters or symbols often related to geographic locations or names

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

FACE

-0.76

 marrow

-0.72

OW

-0.64

 Coliseum

-0.63

hibited

-0.63

OUS

-0.63

IRD

-0.62

Reviewer

-0.62

DVD

-0.61

 BlackBerry

-0.61

POSITIVE LOGITS

ÅŁ

1.41

oÄŁ

1.14

oÄŁan

1.08

Ä±

1.02

nown

0.97

istani

0.96

pai

0.90

edIn

0.86

terday

0.85

 Tayyip

0.84

Activations Density 0.005%

the word "ÅŁ" with various levels of activation strength

specific characters or symbols often related to geographic locations or names

No Comments

No Known Activations