INDEX

Explanations

legal or technical terms related to errors or flaws

oai_token-act-pair · gpt-3.5-turbo

terms related to defects and errors

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

 Lans

-0.73

ahon

-0.67

verson

-0.66

corn

-0.66

 Cornell

-0.66

alach

-0.65

zer

-0.64

oola

-0.63

zees

-0.62

brance

-0.61

POSITIVE LOGITS

 defect

0.99

 defects

0.96

inia

0.89

ively

0.85

itives

0.82

ees

0.81

smanship

0.80

ions

0.80

lect

0.77

lings

0.76

Activations Density 0.007%

legal or technical terms related to errors or flaws

terms related to defects and errors

No Comments

No Known Activations