INDEX

Explanations

mentions of and references to people

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

tnc

-0.87

Accessory

-0.78

Operation

-0.73

COMPLE

-0.70

\/\/

-0.70

catentry

-0.66

 srfAttach

-0.64

Va

-0.63

ory

-0.62

ories

-0.61

POSITIVE LOGITS

who

1.24

 else

1.22

else

1.01

 smugglers

1.00

hood

0.99

 whom

0.98

folk

0.96

who

0.94

 whose

0.93

uscript

0.89

Activations Density 2.766%

mentions of and references to people

No Comments

No Known Activations