INDEX

Explanations

proper names, particularly those involving the term "Doe."

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

specific names, likely with a consistent format like "Jane Doe" followed by a number

oai_token-act-pair · gpt-3.5-turbo

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

OIL

-0.82

ãĥķãĤ©

-0.79

ãĤ·ãĥ£

-0.75

ropolitan

-0.74

cius

-0.72

arching

-0.68

ulated

-0.68

izations

-0.68

Edited

-0.68

mingham

-0.67

POSITIVE LOGITS

Doe

1.20

zie

0.83

hood

0.71

ele

0.70

Pry

0.65

ston

0.64

 Swim

0.64

Gum

0.64

 accuser

0.64

sworth

0.63

Activations Density 0.015%

proper names, particularly those involving the term "Doe."

specific names, likely with a consistent format like "Jane Doe" followed by a number

No Comments

No Known Activations