INDEX

Explanations

proper nouns, likely related to individuals, companies, or products

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

 disg

-0.89

 dart

-0.87

 muster

-0.80

anton

-0.80

 gastro

-0.80

 skilled

-0.79

 gren

-0.77

 popul

-0.75

 versa

-0.75

 forestry

-0.75

POSITIVE LOGITS

 Relationship

1.36

 Problems

1.34

 Away

1.31

 Guys

1.30

 Them

1.30

 Names

1.29

 Experience

1.29

 Replacement

1.29

 Month

1.28

 Difference

1.28

Activations Density 4.422%

proper nouns, likely related to individuals, companies, or products

No Comments

No Known Activations