INDEX

Explanations

actions or events involving public disclosure or communication

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

 Wikimedia

-0.76

Pg

-0.64

gra

-0.60

 Chau

-0.57

 quot

-0.56

"}

-0.55

inf

-0.54

ibal

-0.53

atre

-0.52

hift

-0.52

POSITIVE LOGITS

DERR

0.91

why

0.78

how

0.76

 beforehand

0.73

why

0.72

 orally

0.69

whether

0.68

»Ĵ

0.67

 whether

0.65

cerning

0.64

Activations Density 5.098%

actions or events involving public disclosure or communication

No Comments

No Known Activations