INDEX

Explanations

words related to laws, crimes, and punishments

oai_token-act-pair · gpt-3.5-turbo

terms related to legal penalties and offenses

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

romy

-0.72

iance

-0.72

otin

-0.71

ynthesis

-0.70

hole

-0.69

umbn

-0.68

hall

-0.68

eds

-0.67

ouf

-0.67

por

-0.66

POSITIVE LOGITS

 punishable

1.40

 punished

0.95

 punish

0.92

 unfocusedRange

0.91

 punishment

0.86

 notor

0.84

 Penalty

0.83

 lashes

0.79

 penalties

0.77

 punishments

0.77

Activations Density 0.014%

words related to laws, crimes, and punishments

terms related to legal penalties and offenses

No Comments

No Known Activations