INDEX

Explanations

phrases related to direction and movement

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

 Abyss

-0.65

 Kahn

-0.63

 understatement

-0.62

 Emanuel

-0.59

=]

-0.59

Maker

-0.56

 CLSID

-0.56

gemony

-0.56

 diseng

-0.56

jah

-0.55

POSITIVE LOGITS

stairs

1.16

dates

1.03

river

1.00

grading

1.00

grades

0.97

rights

0.95

dating

0.94

votes

0.92

raised

0.88

graded

0.88

Activations Density 2.726%

phrases related to direction and movement

No Comments

No Known Activations