INDEX

Explanations

words related to safety, security, and situations that are not going well

oai_token-act-pair · gpt-3.5-turbo

phrases indicating safety and well-being

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 wrongful

-0.77

 plagiar

-0.75

grand

-0.73

 Insp

-0.71

 Giant

-0.69

interstitial

-0.68

 Cove

-0.67

 Represent

-0.67

 auctions

-0.67

 parody

-0.66

POSITIVE LOGITS

 calmed

1.21

 resumed

1.13

 awake

1.03

 evacuate

1.02

 calm

1.02

 evacuated

1.02

 alright

0.99

 cleared

0.97

 stabilized

0.95

 ready

0.94

Activations Density 1.038%

words related to safety, security, and situations that are not going well

phrases indicating safety and well-being

No Comments

No Known Activations