INDEX

Explanations

terms related to systems, policies, and guidelines

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

deen

-0.79

tein

-0.75

azar

-0.71

joy

-0.69

worthy

-0.68

ãĥ©ãĥ³

-0.66

igi

-0.66

azines

-0.65

 Stain

-0.65

ighters

-0.63

POSITIVE LOGITS

ologies

1.21

 whereby

1.10

 devised

1.03

 governing

0.97

atically

0.93

 implemented

0.91

 pioneered

0.89

atics

0.89

 underpin

0.85

 dictates

0.85

Activations Density 5.339%

terms related to systems, policies, and guidelines

No Comments

No Known Activations