INDEX

Explanations

words related to deception or manipulation

oai_token-act-pair · gpt-3.5-turbo

references to deceptive practices or tactics used in political or social contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

kens

-0.71

 Horizons

-0.70

 Wander

-0.66

 Parables

-0.65

 CARE

-0.63

anse

-0.60

 aging

-0.60

accompan

-0.60

 experiences

-0.60

inguished

-0.59

POSITIVE LOGITS

 ploy

1.10

 perpetrated

1.10

 blackmail

1.09

 against

1.03

 tactic

1.02

 concoct

1.02

 orchestrated

1.01

 pretext

1.00

 extortion

1.00

 tactics

0.98

Activations Density 0.391%

words related to deception or manipulation

references to deceptive practices or tactics used in political or social contexts

No Comments

No Known Activations