INDEX

Explanations

words related to negative behavior such as abusive, rude, derogatory, and hateful

oai_token-act-pair · gpt-3.5-turbo

language related to abusive or harmful behavior

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

oleon

-0.93

obyl

-0.92

zzo

-0.88

DragonMagazine

-0.85

igham

-0.85

Downloadha

-0.84

zig

-0.83

ortal

-0.83

ariat

-0.82

akeru

-0.82

POSITIVE LOGITS

 behav

1.06

 behaviour

0.94

 abusive

0.86

 behavior

0.85

soever

0.83

 alien

0.82

 distractions

0.80

 aspects

0.80

 undermin

0.79

 slurs

0.78

Activations Density 0.032%

words related to negative behavior such as abusive, rude, derogatory, and hateful

language related to abusive or harmful behavior

No Comments

No Known Activations