INDEX

Explanations

specific locations or proper nouns

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

Lv

-0.67

Dialog

-0.66

_(

-0.64

 Scient

-0.63

 Siber

-0.61

Quantity

-0.58

arbon

-0.58

?,

-0.57

Redditor

-0.57

 Izan

-0.55

POSITIVE LOGITS

 have

0.98

 were

0.98

are

0.96

 rejoice

0.95

 reacted

0.94

 contend

0.92

 insist

0.91

 argue

0.89

 weren

0.83

 fared

0.82

Activations Density 0.421%

specific locations or proper nouns

No Comments

No Known Activations