INDEX

Explanations

references to unique or unconventional ideas or approaches

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

caster

-0.66

Donnell

-0.65

dar

-0.64

casters

-0.64

moon

-0.60

%%

-0.59

$$

-0.58

CLASSIFIED

-0.58

 Choice

-0.58

lain

-0.58

POSITIVE LOGITS

 accordance

1.38

 unison

1.27

clusively

1.25

humane

1.25

 situ

1.21

appropriately

1.17

 earnest

1.17

 lieu

1.16

 terms

1.16

 order

1.14

Activations Density 0.339%

references to unique or unconventional ideas or approaches

No Comments

No Known Activations