INDEX

Explanations

numbers being counted or tallied

oai_token-act-pair · gpt-3.5-turbo

instances of the word "counting" and related terms

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

cel

-0.76

ggie

-0.68

por

-0.66

 cover

-0.65

imar

-0.64

 Correspond

-0.64

ieties

-0.64

ocker

-0.61

acia

-0.61

 Charter

-0.61

POSITIVE LOGITS

enance

1.46

stakes

0.89

 counting

0.88

ENCY

0.86

downs

0.86

reys

0.79

 counted

0.78

esses

0.77

 bells

0.75

ARDS

0.73

Activations Density 0.009%

numbers being counted or tallied

instances of the word "counting" and related terms

No Comments

No Known Activations