INDEX

Explanations

proper nouns related to people or places, specifically 'Mills'

oai_token-act-pair · gpt-3.5-turbo

references to individuals or entities associated with "Mills" and related terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

liest

-0.73

ctory

-0.72

 Falk

-0.71

esses

-0.67

 stern

-0.66

gers

-0.65

ctica

-0.64

 LAPD

-0.64

 Saddam

-0.63

 fierce

-0.63

POSITIVE LOGITS

pora

1.07

 Mills

0.99

essage

0.97

creen

0.86

onduct

0.84

hirt

0.84

boro

0.80

mallow

0.80

arty

0.79

onductor

0.78

Activations Density 0.016%

proper nouns related to people or places, specifically 'Mills'

references to individuals or entities associated with "Mills" and related terms

No Comments

No Known Activations