INDEX

Explanations

the word "type" followed by either a number or a word in the context of different topics such as medical conditions, geographic features, and Pokémon characteristics

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

 Rodrigo

-0.65

nesday

-0.65

 Zimmer

-0.63

uthor

-0.63

 Bots

-0.62

 Morning

-0.61

iae

-0.61

ahar

-0.61

aeper

-0.60

pload

-0.60

POSITIVE LOGITS

face

1.44

faces

1.33

etting

1.13

ahead

1.09

casting

1.03

etter

0.94

alias

0.84

classes

0.84

 locality

0.80

cast

0.75

Activations Density 0.535%

the word "type" followed by either a number or a word in the context of different topics such as medical conditions, geographic features, and Pokémon characteristics

No Comments

No Known Activations