INDEX

Explanations

phrases or sentences indicating sarcasm or jest

oai_token-act-pair · gpt-3.5-turbo

phrases related to teasing or joking

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

marked

-0.89

ŃĶ

-0.85

por

-0.75

marks

-0.72

porter

-0.70

gars

-0.68

actions

-0.67

Decre

-0.67

lining

-0.66

morph

-0.66

POSITIVE LOGITS

 kidding

1.25

 aside

0.88

 joking

0.79

 aloud

0.74

 spared

0.69

 hypoc

0.66

renheit

0.65

 posit

0.64

 Goodbye

0.62

 jokes

0.62

Activations Density 0.007%

phrases or sentences indicating sarcasm or jest

phrases related to teasing or joking

No Comments

No Known Activations