INDEX

Explanations

the word "liking" or related forms at high activation levels

oai_token-act-pair · gpt-3.5-turbo

words and phrases that indicate approval or preference

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

arta

-0.91

inas

-0.81

chin

-0.80

nex

-0.79

cise

-0.72

inar

-0.71

cci

-0.70

strip

-0.68

src

-0.68

inary

-0.68

POSITIVE LOGITS

lihood

1.55

0.91

lier

0.86

ably

0.81

liest

0.80

joy

0.72

 Muss

0.71

Pwr

0.70

liness

0.69

ability

0.67

Activations Density 0.037%

the word "liking" or related forms at high activation levels

words and phrases that indicate approval or preference

No Comments

No Known Activations