INDEX

Explanations

words related to legal actions, particularly pleas of guilt

oai_token-act-pair · gpt-3.5-turbo

phrases indicating legal guilt or guilty pleas

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

adish

-0.75

edia

-0.71

dos

-0.67

clips

-0.65

osi

-0.65

dust

-0.64

ffee

-0.63

hot

-0.62

 vantage

-0.62

ãĤ¡

-0.61

POSITIVE LOGITS

 plea

0.96

 Guilty

0.92

 guilty

0.89

 verdict

0.87

 manslaughter

0.85

 voluntarily

0.82

 indict

0.80

Prosecut

0.80

 probation

0.78

icts

0.78

Activations Density 0.033%

words related to legal actions, particularly pleas of guilt

phrases indicating legal guilt or guilty pleas

No Comments

No Known Activations