INDEX

Explanations

proper nouns or titles of authority figures

oai_token-act-pair · gpt-3.5-turbo

instances of the word "top" followed by a ranking or status

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ufact

-0.70

 Consent

-0.68

 Gaul

-0.67

 Griffith

-0.64

 Anth

-0.63

 Manual

-0.63

cci

-0.62

 Advent

-0.62

 Arri

-0.62

 Kath

-0.61

POSITIVE LOGITS

most

1.08

ographical

1.02

 tier

1.01

ography

0.88

ographically

0.88

 notch

0.87

 drawer

0.85

 scorer

0.85

eka

0.84

ranking

0.83

Activations Density 0.035%

proper nouns or titles of authority figures

instances of the word "top" followed by a ranking or status

No Comments

No Known Activations