INDEX

Explanations

the verb "wrote"

oai_token-act-pair · gpt-3.5-turbo

instances of the word "wrote" indicating discussions about authorship or writing

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

Ĭ±

-0.88

azar

-0.72

Enlarge

-0.69

alon

-0.69

 Unsure

-0.69

illon

-0.68

xon

-0.67

allows

-0.67

nel

-0.67

imation

-0.66

POSITIVE LOGITS

 poems

0.97

 essays

0.88

acters

0.87

smanship

0.86

writer

0.83

 poem

0.82

writers

0.81

 poetry

0.81

itatively

0.80

 notebooks

0.78

Activations Density 0.045%

the verb "wrote"

instances of the word "wrote" indicating discussions about authorship or writing

No Comments

No Known Activations