INDEX

Explanations

capitalized words or proper nouns with the prefix "ar-" followed by a lowercase word

oai_token-act-pair · gpt-3.5-turbo

occurrences of the substring "ar" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

å§«

-0.89

éĹĺ

-0.70

 Wilmington

-0.69

 Seller

-0.69

 Vide

-0.69

ERSON

-0.67

STER

-0.66

eters

-0.65

wich

-0.65

ega

-0.64

POSITIVE LOGITS

beit

1.08

thur

1.06

duino

1.01

ar

0.89

ificial

0.84

atars

0.83

ithmetic

0.82

gebra

0.82

ctic

0.81

TeX

0.81

Activations Density 0.004%

capitalized words or proper nouns with the prefix "ar-" followed by a lowercase word

occurrences of the substring "ar" in various contexts

No Comments

No Known Activations