INDEX

Explanations

numerical values related to quantity or comparison

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_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

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

ĸļ

-0.80

merce

-0.72

inka

-0.68

aimon

-0.67

warts

-0.67

selves

-0.65

bard

-0.64

 Ended

-0.64

emon

-0.64

çīĪ

-0.64

POSITIVE LOGITS

 allowable

1.02

of

0.97

 difference

0.86

 thereof

0.83

liest

0.79

 deviation

0.73

 payable

0.73

 permissible

0.73

 nearest

0.73

of

0.72

Activations Density 0.263%

numerical values related to quantity or comparison

No Comments

No Known Activations