INDEX

Explanations

geographic locations, particularly focused on Northern regions

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

inger

-0.83

icit

-0.79

ign

-0.78

lest

-0.75

bs

-0.73

aday

-0.73

enable

-0.72

bley

-0.72

ifice

-0.72

andal

-0.72

POSITIVE LOGITS

 Hemisphere

1.33

 hemisphere

1.28

most

1.07

 Isles

0.94

 regions

0.91

 provinces

0.90

 Ontario

0.89

 Illinois

0.88

 Province

0.88

 California

0.84

Activations Density 0.581%

geographic locations, particularly focused on Northern regions

No Comments

No Known Activations