INDEX

Explanations

proper names or surnames, especially ending in "lis."

oai_token-act-pair · gpt-3.5-turbo

references to specific individuals or names

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

esa

-0.75

¥ŀ

-0.72

ħĭ

-0.71

ed

-0.69

į

-0.69

 Pref

-0.68

ozy

-0.67

ortunately

-0.65

llah

-0.64

ģ

-0.64

POSITIVE LOGITS

berger

1.05

ovie

0.95

orescence

0.87

boa

0.87

ãĥ³ãĤ¸

0.83

backer

0.78

uminati

0.76

tesy

0.76

toe

0.75

bourg

0.75

Activations Density 0.032%

proper names or surnames, especially ending in "lis."

references to specific individuals or names

No Comments

No Known Activations