INDEX

Explanations

phrases that contrast different sides or perspectives on a topic

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

theless

-0.67

zon

-0.61

 RELEASE

-0.60

Effective

-0.60

Progress

-0.60

cit

-0.60

 Updated

-0.59

burg

-0.57

zan

-0.56

anned

-0.55

POSITIVE LOGITS

 side

1.22

worldly

1.20

 hemisphere

0.93

 hand

0.92

 half

0.89

most

0.88

iest

0.83

 Side

0.83

end

0.83

 halves

0.81

Activations Density 0.629%

phrases that contrast different sides or perspectives on a topic

No Comments

No Known Activations