INDEX

Explanations

expressions conveying positive emotions or exclamations of surprise

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "goodness" in various contexts

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

vent

-0.75

 Mast

-0.71

pent

-0.71

rik

-0.69

ZI

-0.66

ribe

-0.65

 Miner

-0.65

bi

-0.63

 Veteran

-0.63

 Psychiatric

-0.62

POSITIVE LOGITS

 goodness

1.53

 kindness

0.95

iness

0.91

 gracious

0.90

owship

0.87

istries

0.85

 generosity

0.85

liness

0.84

 tremend

0.81

 goodies

0.81

Activations Density 0.006%

expressions conveying positive emotions or exclamations of surprise

references to the concept of "goodness" in various contexts

No Comments

No Known Activations