INDEX

Explanations

numerical lists like countdowns or ranges

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

Hoo

-0.70

 Richards

-0.67

 Lanka

-0.67

 Proc

-0.66

 Sund

-0.65

 Borders

-0.64

 Clyde

-0.63

Ins

-0.62

 Cassidy

-0.62

 Ober

-0.61

POSITIVE LOGITS

year

1.48

minute

1.35

sided

1.34

month

1.34

legged

1.30

hour

1.29

digit

1.28

dimensional

1.25

degree

1.25

point

1.21

Activations Density 0.668%

numerical lists like countdowns or ranges

No Comments

No Known Activations