INDEX

Explanations

keywords or symbols used for emphasis or headings in online content

oai_token-act-pair · gpt-3.5-turbo

instances of asterisks or special symbols used for emphasis or decoration

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

',"

-0.74

',

-0.74

"—

-0.70

erc

-0.63

ERC

-0.63

'?

-0.63

'-

-0.62

-0.59

aber

-0.59

":

-0.58

POSITIVE LOGITS

***

3.21

**

3.11

 ****

2.62

 ********************************

1.79

1.72

 Â«

1.64

+++

1.61

###

1.60

*)

1.51

##

1.45

Activations Density 0.013%

keywords or symbols used for emphasis or headings in online content

instances of asterisks or special symbols used for emphasis or decoration

No Comments

No Known Activations