INDEX

Explanations

mentions of specific dates in a variety of contexts

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

 swearing

-0.63

awaru

-0.59

 Lumpur

-0.57

 fencing

-0.56

 mascul

-0.56

Wag

-0.55

 Varg

-0.55

 Sweeney

-0.55

Vog

-0.55

 vested

-0.54

POSITIVE LOGITS

Â½

1.26

1.23

Password

1.22

1.20

1.18

1.08

st

1.05

1.02

nd

1.00

Activations Density 1.558%

mentions of specific dates in a variety of contexts

No Comments

No Known Activations