INDEX

Explanations

references to various pop culture references, political topics, internet links, and cultural commentary

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

 Barrier

-0.67

 seiz

-0.66

 Downs

-0.63

 Phant

-0.62

OTOS

-0.62

 Sandwich

-0.62

 scatter

-0.62

 shroud

-0.61

 Slug

-0.61

 Shroud

-0.61

POSITIVE LOGITS

º

0.90

¬

0.86

£

0.85

then

0.84

ĸļ

0.83

Ń

0.83

¡

0.81

ķ

0.81

especially

0.81

Ĵ

0.78

Activations Density 0.128%

references to various pop culture references, political topics, internet links, and cultural commentary

No Comments

No Known Activations