INDEX

Explanations

phrases related to expressing opinions or arguments

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ĳå£«

-0.76

 Peaks

-0.65

ãĤ´ãĥ³

-0.64

 Kers

-0.60

 Viper

-0.56

 sole

-0.56

 widest

-0.54

 majority

-0.53

bye

-0.53

 rogue

-0.52

POSITIVE LOGITS

 than

1.03

than

1.03

erous

0.98

minecraft

0.81

culated

0.78

 Than

0.78

worldly

0.76

efficient

0.73

uries

0.70

itional

0.69

Activations Density 0.123%

phrases related to expressing opinions or arguments

No Comments

No Known Activations