INDEX

Explanations

mentions of specific names or entities ending in 'nan'

oai_token-act-pair · gpt-3.5-turbo

instances of the term "nan" and variations of the word related to numerical data

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

ACTED

-0.76

UID

-0.70

ãģį

-0.65

IRD

-0.64

 Caribbean

-0.62

 value

-0.62

ENGTH

-0.62

 Scandinavian

-0.61

 Dangerous

-0.61

 frightening

-0.61

POSITIVE LOGITS

nan

1.58

igans

1.24

thood

1.07

apixel

1.06

ufact

1.05

ning

0.99

ascus

0.92

uclear

0.92

wu

0.92

apolis

0.91

Activations Density 0.004%

mentions of specific names or entities ending in 'nan'

instances of the term "nan" and variations of the word related to numerical data

No Comments

No Known Activations