INDEX

Explanations

the word "absolutely" with high activation values

oai_token-act-pair · gpt-3.5-turbo

strong affirmations or positive assertions

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

coh

-0.69

 Skydragon

-0.66

 mole

-0.66

squ

-0.64

é¾

-0.64

 creeping

-0.63

 quarters

-0.63

æł

-0.62

oresc

-0.62

Ó

-0.61

POSITIVE LOGITS

Absolutely

0.86

ogether

0.85

Mine

0.79

True

0.78

 Absolutely

0.78

 Wrong

0.75

Yes

0.74

 Definitely

0.72

 Possibly

0.72

leans

0.71

Activations Density 0.023%

the word "absolutely" with high activation values

strong affirmations or positive assertions

No Comments

No Known Activations