INDEX

Explanations

objects or concepts being ignored or dismissed

oai_token-act-pair · gpt-3.5-turbo

the word "ign" in various contexts and forms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Springfield

-0.67

------------------------------------------------

-0.65

HAHAHAHA

-0.64

âĨĳ

-0.63

 Akin

-0.63

 Spears

-0.63

 Heard

-0.62

 Safari

-0.62

EStream

-0.62

????????

-0.62

POSITIVE LOGITS

ign

1.22

iew

1.10

ificant

1.01

ific

1.01

ame

0.98

idad

0.96

antly

0.95

igible

0.92

olor

0.91

ibrary

0.91

Activations Density 0.008%

objects or concepts being ignored or dismissed

the word "ign" in various contexts and forms

No Comments

No Known Activations