INDEX

Explanations

phrases related to political and historical events

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

emia

-0.74

externalActionCode

-0.69

ahime

-0.66

illary

-0.66

achev

-0.65

SG

-0.64

 Brune

-0.64

 antioxid

-0.63

================================================================

-0.61

=-=-=-=-=-=-=-=-

-0.61

POSITIVE LOGITS

theless

1.59

 dreamed

1.14

 heard

0.88

 imagined

0.87

 doubted

0.87

 existed

0.86

 tasted

0.85

entimes

0.84

 bothered

0.80

 wondered

0.79

Activations Density 2.629%

phrases related to political and historical events

No Comments

No Known Activations