INDEX

Explanations

numbers and proper names

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

holders

-0.95

holder

-0.80

 Illum

-0.68

cluding

-0.66

 constit

-0.58

 rooting

-0.57

iple

-0.57

 carriers

-0.56

CONCLUS

-0.56

houses

-0.56

POSITIVE LOGITS

upiter

1.20

igsaw

1.19

ournals

1.19

ealous

1.17

unction

1.16

umbo

1.09

oint

1.08

utsu

1.07

itsu

1.04

acket

1.02

Activations Density 3.954%

numbers and proper names

No Comments

No Known Activations