INDEX

Explanations

phrases related to attempting, trying, arguing, and explaining

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 Used

-0.70

faces

-0.65

 Falling

-0.64

 Legs

-0.63

Flo

-0.61

 Printed

-0.61

minus

-0.61

 Lots

-0.61

 Plain

-0.60

 Sitting

-0.59

POSITIVE LOGITS

 emulate

1.44

 recreate

1.41

 revive

1.38

 convince

1.36

 minimize

1.36

 reconcile

1.31

 capitalize

1.30

 improve

1.30

 replicate

1.29

 educate

1.28

Activations Density 1.528%

phrases related to attempting, trying, arguing, and explaining

No Comments

No Known Activations