INDEX

Explanations

names or proper nouns

oai_token-act-pair · gpt-3.5-turbo

suffixes or fragments of words

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Stella

-0.63

 Paula

-0.60

Nar

-0.59

 Vengeance

-0.58

 faire

-0.57

BUR

-0.54

 Psycho

-0.54

 Lily

-0.54

CAP

-0.54

Pix

-0.53

POSITIVE LOGITS

enegger

1.12

 himself

1.04

 testified

0.98

 oversaw

0.90

 wrote

0.88

's

0.87

 penned

0.87

baum

0.85

 told

0.85

 specializes

0.84

Activations Density 0.166%

names or proper nouns

suffixes or fragments of words

No Comments

No Known Activations