INDEX

Explanations

numerical quantities expressed in words, particularly thousands and hundreds

oai_token-act-pair · gpt-3.5-turbo

numerical values, especially those related to quantities and measurements

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Goldberg

-0.70

 Cinderella

-0.70

mania

-0.69

 Barkley

-0.68

 fixture

-0.67

 cracking

-0.65

Iv

-0.64

cre

-0.62

 Chamberlain

-0.62

 whipping

-0.61

POSITIVE LOGITS

ousand

3.53

undred

2.89

alian

1.94

ousands

1.77

 Thousand

1.64

undreds

1.41

irty

1.27

aughtered

1.19

irteen

1.19

aughters

1.08

Activations Density 0.059%

numerical quantities expressed in words, particularly thousands and hundreds

numerical values, especially those related to quantities and measurements

No Comments

No Known Activations