INDEX

Explanations

information related to technical details, services, and instructions

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

osen

-0.74

ese

-0.73

apes

-0.72

alty

-0.70

zz

-0.69

shr

-0.69

ugh

-0.67

asar

-0.66

å¥³

-0.66

acks

-0.65

POSITIVE LOGITS

 upon

1.01

 loosely

0.89

 solely

0.86

uates

0.78

onial

0.77

plates

0.72

atively

0.71

uate

0.68

 principally

0.68

awaru

0.66

Activations Density 2.053%

information related to technical details, services, and instructions

No Comments

No Known Activations