INDEX

Explanations

references to abuse and misuse of power or authority

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-att_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-attn-out-SAEs/v5_32k_layer_3

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.3.hook_attn_out

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ufact

-0.86

eele

-0.84

ritz

-0.74

travel

-0.73

eus

-0.72

udeau

-0.71

ppa

-0.71

ften

-0.71

UTERS

-0.70

apixel

-0.70

POSITIVE LOGITS

 medication

0.74

 medications

0.73

 narc

0.66

 substances

0.65

 opioids

0.64

 treatment

0.63

alcohol

0.63

 meth

0.62

 stigma

0.62

 hormone

0.62

Activations Density 0.170%

references to abuse and misuse of power or authority

No Comments

No Known Activations