INDEX

Explanations

instances of the word "smart" and related concepts of intelligence

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

ĸļ

-0.79

artifacts

-0.74

 dissatisf

-0.68

Reloaded

-0.66

 Divinity

-0.64

 Syndrome

-0.62

avored

-0.62

ãĥĺãĥ©

-0.60

riott

-0.60

 stret

-0.59

POSITIVE LOGITS

guy

0.97

ctl

0.93

sonian

0.93

ness

0.90

ly

0.90

arse

0.89

ass

0.88

ling

0.86

found

0.86

asses

0.84

Activations Density 0.013%

instances of the word "smart" and related concepts of intelligence

No Comments

No Known Activations