INDEX

Explanations

decisions that have been supported or maintained by higher authorities

oai_token-act-pair · gpt-3.5-turbo

terms related to legal affirmations or support for rulings

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Plot

-0.82

Psy

-0.75

Bul

-0.75

di

-0.73

 magnet

-0.67

 panic

-0.66

Bus

-0.64

 hypnot

-0.63

 Tiny

-0.63

Jet

-0.62

POSITIVE LOGITS

 uphold

2.97

 upheld

2.62

 upholding

2.32

 overturned

1.74

uph

1.47

 affirmed

1.37

 overturn

1.36

 reinstated

1.23

rul

1.20

 abide

1.18

Activations Density 0.025%

decisions that have been supported or maintained by higher authorities

terms related to legal affirmations or support for rulings

No Comments

No Known Activations