INDEX

Explanations

numbers with 'minutes', 'days', or 'years'

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

unks

-0.78

yz

-0.74

urat

-0.67

Nay

-0.66

 Azerb

-0.66

olls

-0.64

ibi

-0.64

 Qiao

-0.64

ulhu

-0.62

 Pengu

-0.62

POSITIVE LOGITS

necess

0.87

coll

0.86

relations

0.84

communications

0.83

under

0.81

environment

0.81

administ

0.81

communication

0.81

tem

0.80

recent

0.80

Activations Density 0.008%

numbers with 'minutes', 'days', or 'years'

No Comments

No Known Activations