INDEX

Explanations

information related to items or objects, possibly focusing on craftsmanship or design details

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

Politics

-0.99

arcity

-0.98

Economic

-0.92

Services

-0.88

Financial

-0.88

ivals

-0.88

enario

-0.85

GET

-0.83

Events

-0.83

humans

-0.81

POSITIVE LOGITS

 protr

1.78

 diameter

1.38

 curved

1.36

 molded

1.35

 leather

1.32

 translucent

1.32

 thickness

1.31

 rectangular

1.31

 sleeves

1.29

 oval

1.27

Activations Density 7.389%

information related to items or objects, possibly focusing on craftsmanship or design details

No Comments

No Known Activations