INDEX

Explanations

phrases related to legal and political issues

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

•Test list

"with"

•"with" features in 768 and 1536

•Canonical Fig 12

No Comments

Negative Logits

sburg

-0.67

fair

-0.67

sov

-0.66

NN

-0.66

sat

-0.64

BF

-0.63

BP

-0.63

hops

-0.61

blast

-0.60

pad

-0.60

POSITIVE LOGITS

 regards

1.65

stood

1.64

 regard

1.62

draw

1.54

drawn

1.44

 impunity

1.40

 respect

1.30

standing

1.25

holding

1.20

 utmost

0.94

Activations Density 1.392%

phrases related to legal and political issues

No Comments

No Known Activations