INDEX

Explanations

words related to locations or organizations

oai_token-act-pair · gpt-3.5-turbo

references to specific podcasts or film titles

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

ILY

-0.69

inventoryQuantity

-0.69

 acknow

-0.68

 positively

-0.68

pi

-0.66

 benef

-0.65

 downhill

-0.65

ably

-0.63

è¦ļéĨĴ

-0.62

 compr

-0.62

POSITIVE LOGITS

 Doors

0.94

 Roads

0.89

lander

0.88

 Things

0.85

erton

0.84

ward

0.83

 Limits

0.82

 Lives

0.81

brook

0.81

 Deck

0.81

Activations Density 0.133%

words related to locations or organizations

references to specific podcasts or film titles

No Comments

No Known Activations