INDEX

Explanations

dates in the format Month followed by a number

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ļéĨĴ

-0.80

*/(

-0.71

pn

-0.71

swick

-0.69

ħĭ

-0.68

 CODE

-0.66

PDATE

-0.66

 tray

-0.65

culosis

-0.65

estern

-0.63

POSITIVE LOGITS

nard

1.43

flower

1.28

onna

1.20

fair

1.19

hew

1.09

oral

1.05

weather

1.04

berry

0.95

bach

0.90

umi

0.89

Activations Density 0.031%

dates in the format Month followed by a number

No Comments

No Known Activations