INDEX

Explanations

titles of publications or works

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

ONSORED

-0.79

xual

-0.77

cius

-0.72

OPA

-0.72

aminer

-0.71

acebook

-0.69

ï¸ı

-0.69

osure

-0.68

ashtra

-0.67

ouse

-0.66

POSITIVE LOGITS

bies

1.45

bie

1.45

 Zealand

1.22

foundland

1.16

 arrivals

1.16

 Orleans

1.10

 York

1.07

 additions

0.99

 YORK

0.94

 Hampshire

0.91

Activations Density 2.472%

titles of publications or works

No Comments

No Known Activations