INDEX

Explanations

numbers and measurements with the word "roughly" preceding them

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate approximate quantities or measures

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

oli

-0.72

actic

-0.65

arella

-0.63

 enthus

-0.61

era

-0.61

oleon

-0.60

 Definitely

-0.58

ves

-0.58

oji

-0.58

 decidedly

-0.57

POSITIVE LOGITS

 equivalent

1.00

 analogous

0.95

 eighty

0.91

0.86

 seventy

0.86

 equal

0.85

0.84

0.83

Activations Density 0.026%

numbers and measurements with the word "roughly" preceding them

phrases that indicate approximate quantities or measures

No Comments

No Known Activations