INDEX

Explanations

words related to the beginning or initiation of actions or events

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

aths

-0.68

obi

-0.68

athed

-0.66

acho

-0.66

ugs

-0.66

rero

-0.63

owl

-0.63

itsch

-0.62

 entirety

-0.62

warts

-0.59

POSITIVE LOGITS

 anew

1.37

nings

0.94

 raining

0.81

 behaving

0.80

 hostilities

0.74

rek

0.73

 airing

0.72

 bothering

0.70

 unravel

0.70

 worrying

0.70

Activations Density 2.865%

words related to the beginning or initiation of actions or events

No Comments

No Known Activations