INDEX

Explanations

time periods mentioned in years or days

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

âĨĳ

-0.73

Tomorrow

-0.69

cci

-0.67

odder

-0.67

behind

-0.67

Inside

-0.66

Behind

-0.66

enf

-0.66

osite

-0.63

 Ahead

-0.62

POSITIVE LOGITS

 period

1.30

 span

1.24

 millennia

1.06

period

1.01

 timeframe

0.99

 decades

0.98

 centuries

0.93

 lifespan

0.93

 periods

0.92

decade

0.91

Activations Density 0.105%

time periods mentioned in years or days

No Comments

No Known Activations