INDEX

Explanations

phrases related to entering contests or giveaways

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

 Archangel

-0.68

 gunned

-0.60

 trust

-0.60

 backdrop

-0.60

heet

-0.59

ished

-0.58

utsche

-0.58

 shepherd

-0.57

cks

-0.56

ector

-0.56

POSITIVE LOGITS

prising

1.51

prise

1.39

prises

1.37

tainment

1.25

tain

1.24

tained

1.09

TAIN

1.09

taining

1.08

 INTO

1.02

 into

0.92

Activations Density 0.538%

phrases related to entering contests or giveaways

No Comments

No Known Activations