INDEX

Explanations

numerical measurements such as scores, statistics, or other quantifiable data

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

atem

-0.55

 Uriel

-0.55

 Sorceress

-0.54

Aly

-0.53

£ı

-0.53

 retweet

-0.53

 Jagu

-0.53

 DRAG

-0.51

 Lara

-0.51

 Levant

-0.51

POSITIVE LOGITS

th

1.43

1.27

1.24

1.21

1.20

1.19

Activations Density 1.377%

numerical measurements such as scores, statistics, or other quantifiable data

No Comments

No Known Activations