INDEX

Explanations

proper nouns related to political figures, government, and official titles

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

MpServer

-0.80

ratom

-0.74

essen

-0.69

Î³

-0.67

alle

-0.64

lot

-0.63

lighting

-0.61

Medium

-0.61

trace

-0.61

Alien

-0.60

POSITIVE LOGITS

 Barack

1.11

clinton

1.07

 Obama

0.97

ially

0.97

 Lyndon

0.94

 Jinping

0.92

 Recep

0.84

ial

0.83

 hopeful

0.82

 candidate

0.80

Activations Density 2.821%

proper nouns related to political figures, government, and official titles

No Comments

No Known Activations