INDEX

Explanations

references to fortifications or places considered secure or heavily protected

oai_token-act-pair · gpt-3.5-turbo

terms associated with fortified positions or strong defensive locations

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

hy

-0.83

ordan

-0.82

andro

-0.78

respond

-0.75

ivan

-0.74

apers

-0.74

matter

-0.73

ensation

-0.71

videos

-0.70

uci

-0.70

POSITIVE LOGITS

 stronghold

1.41

 strongh

1.32

 bast

0.86

 elector

0.82

 advoc

0.81

stead

0.78

urst

0.78

rongh

0.76

 tradem

0.75

 enclave

0.73

Activations Density 0.012%

references to fortifications or places considered secure or heavily protected

terms associated with fortified positions or strong defensive locations

No Comments

No Known Activations