INDEX

Explanations

words related to music albums, facial features, and political figures

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

Data Type

torch.float32

Hook Point

blocks.8.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

ä»£

-0.84

icals

-0.79

igious

-0.76

OPLE

-0.76

ICAL

-0.73

naires

-0.68

Ģ

-0.67

 Lauder

-0.66

ively

-0.65

icism

-0.65

POSITIVE LOGITS

ball

1.01

wark

1.00

oons

0.94

ãĤ§

0.92

ount

0.88

itals

0.87

light

0.87

suit

0.81

lett

0.81

oon

0.80

Activations Density 5.133%

words related to music albums, facial features, and political figures

No Comments

No Known Activations