INDEX

Explanations

words related to political processes and decision-making

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Afee

-0.73

 fortun

-0.68

NK

-0.68

 Falling

-0.67

quartered

-0.65

 Moroc

-0.65

 Kinnikuman

-0.64

CTV

-0.64

iquette

-0.63

 Aires

-0.63

POSITIVE LOGITS

 filibuster

1.44

 approve

1.43

 pass

1.38

 amend

1.35

 repeal

1.32

 override

1.29

 enact

1.27

 veto

1.26

 vote

1.26

 defund

1.25

Activations Density 0.175%

words related to political processes and decision-making

No Comments

No Known Activations