INDEX

Explanations

exaggerated or extreme adjectives

oai_token-act-pair · gpt-3.5-turbo

phrases that convey a sense of near completeness or approximation

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

tein

-0.73

yi

-0.68

ioch

-0.68

alez

-0.66

lest

-0.65

ems

-0.64

iere

-0.64

igans

-0.64

seller

-0.63

aley

-0.63

POSITIVE LOGITS

 unchanged

0.81

 indistinguishable

0.80

etheless

0.79

thood

0.78

 identical

0.76

 unemploy

0.76

 illiter

0.76

electric

0.75

 unheard

0.71

 unint

0.68

Activations Density 0.017%

exaggerated or extreme adjectives

phrases that convey a sense of near completeness or approximation

No Comments

No Known Activations