INDEX

Explanations

words related to low-quality or undesirable items or attributes

oai_token-act-pair · gpt-3.5-turbo

references to "junk" or unhealthy food

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

cing

-0.71

Æ

-0.70

oral

-0.68

phrine

-0.68

sis

-0.68

BRE

-0.67

mberg

-0.67

zza

-0.67

tical

-0.67

 Lauder

-0.66

POSITIVE LOGITS

 heap

0.93

 junk

0.86

glers

0.85

ie

0.85

busters

0.84

ies

0.83

iness

0.82

ãģĦ

0.79

 drawer

0.79

iest

0.78

Activations Density 0.029%

words related to low-quality or undesirable items or attributes

references to "junk" or unhealthy food

No Comments

No Known Activations