INDEX

Explanations

proper nouns related to a particular geographical location

oai_token-act-pair · gpt-3.5-turbo

references to the term "oma," indicating types of tumors or cancer

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

balance

-0.98

anooga

-0.73

BOOK

-0.72

alien

-0.72

leness

-0.70

 intuitive

-0.70

flush

-0.69

ruck

-0.69

ship

-0.67

ellation

-0.67

POSITIVE LOGITS

oma

1.01

pha

0.84

 Nost

0.81

atis

0.79

Naz

0.78

Tos

0.76

ascus

0.74

ography

0.73

 Luxem

0.72

ucc

0.71

Activations Density 0.008%

proper nouns related to a particular geographical location

references to the term "oma," indicating types of tumors or cancer

No Comments

No Known Activations