INDEX

Explanations

adjectives conveying a strong and clear contrast

oai_token-act-pair · gpt-3.5-turbo

the word "stark" and its variations in contexts highlighting contrasts or extremes

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

uthor

-0.72

annis

-0.71

uters

-0.70

ipop

-0.70

 diligently

-0.68

phis

-0.67

PU

-0.67

 safely

-0.66

hops

-0.65

 hemor

-0.64

POSITIVE LOGITS

 contrasts

1.12

ly

1.04

 contrast

1.03

 stark

0.88

 naked

0.82

est

0.81

iary

0.76

 contrasting

0.75

 difference

0.74

er

0.74

Activations Density 0.013%

adjectives conveying a strong and clear contrast

the word "stark" and its variations in contexts highlighting contrasts or extremes

No Comments

No Known Activations