INDEX

Explanations

words related to betrayal or defiance

oai_token-act-pair · gpt-3.5-turbo

references to rogue entities or groups

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

iences

-0.85

raints

-0.83

È

-0.79

akeru

-0.79

ori

-0.79

anza

-0.75

eph

-0.74

lished

-0.74

Wel

-0.74

uffer

-0.73

POSITIVE LOGITS

 rogue

1.11

 thug

0.89

 queen

0.82

esses

0.78

 undermin

0.75

 faction

0.73

 reneg

0.73

 behavior

0.71

 assassin

0.70

 predator

0.70

Activations Density 0.007%

words related to betrayal or defiance

references to rogue entities or groups

No Comments

No Known Activations