INDEX

Explanations

words related to proper names or specific entities

oai_token-act-pair · gpt-3.5-turbo

instances of the letter 'e'

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

atform

-0.56

 fodder

-0.56

 sustainability

-0.53

¶ħ

-0.53

 wallets

-0.53

 Sonny

-0.52

 minds

-0.51

 otherwise

-0.50

 assorted

-0.49

 SHARES

-0.49

POSITIVE LOGITS

3.97

ea

2.01

ei

1.87

eus

1.80

eh

1.74

ele

1.70

ez

1.63

eur

1.63

ean

1.58

eu

1.57

Activations Density 0.029%

words related to proper names or specific entities

instances of the letter 'e'

No Comments

No Known Activations