INDEX

Explanations

instances of the word "per" followed by a number, likely indicating a rate or measurement

oai_token-act-pair · gpt-3.5-turbo

phrases indicating a measurement or quantity expressed as "per" followed by a numeric value

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

brand

-0.72

Je

-0.71

zx

-0.68

ratulations

-0.67

alogy

-0.67

rade

-0.66

Cause

-0.65

hess

-0.64

ese

-0.63

andr

-0.62

POSITIVE LOGITS

per

3.32

per

1.79

Per

1.74

Per

1.59

PER

1.58

 apiece

1.29

 percent

1.17

PER

1.11

pc

1.07

 Percent

0.97

Activations Density 0.036%

instances of the word "per" followed by a number, likely indicating a rate or measurement

phrases indicating a measurement or quantity expressed as "per" followed by a numeric value

No Comments

No Known Activations