INDEX

Explanations

instances where actions are done willingly, especially when referring to providing information or assistance

oai_token-act-pair · gpt-3.5-turbo

terms related to voluntary actions and participation

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 mock

-0.73

 holog

-0.71

 replication

-0.60

 Lann

-0.59

acements

-0.59

 genesis

-0.59

 hilar

-0.59

 replica

-0.59

 illust

-0.58

 spin

-0.58

POSITIVE LOGITS

 voluntarily

3.05

 peacefully

1.96

 willingly

1.87

 unilaterally

1.43

 responsibly

1.26

 anonymously

1.26

 uncond

1.26

 lawfully

1.22

 unlawfully

1.20

untarily

1.11

Activations Density 0.019%

instances where actions are done willingly, especially when referring to providing information or assistance

terms related to voluntary actions and participation

No Comments

No Known Activations