INDEX

Explanations

phrases related to political discourse

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

 GOODMAN

-0.73

adm

-0.70

 diplom

-0.70

 Aval

-0.67

 Rebels

-0.65

 Lauder

-0.64

ģĸ

-0.63

 compe

-0.61

 Exodus

-0.61

 susp

-0.60

POSITIVE LOGITS

ossible

1.31

redict

1.31

aired

1.29

ivot

1.25

ierce

1.24

ardon

1.23

ixels

1.20

oultry

1.20

ulse

1.19

uppet

1.16

Activations Density 0.345%

phrases related to political discourse

No Comments

No Known Activations