INDEX

Explanations

phrases related to negative or controversial actions

oai_token-act-pair · gpt-3.5-turbo

references to actions perceived as deceptive or insincere, often labeled as "stunts" or "gestures"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

izont

-0.71

iaries

-0.71

aido

-0.69

abella

-0.69

inas

-0.68

ousands

-0.68

ancies

-0.66

leys

-0.64

ecycle

-0.64

upt

-0.64

POSITIVE LOGITS

 designed

1.33

 perpetrated

1.31

 aimed

1.29

 intended

1.20

 concoct

1.19

 undertaken

1.18

 borne

1.14

 unworthy

1.13

 devised

1.12

 orchestrated

1.10

Activations Density 0.287%

phrases related to negative or controversial actions

references to actions perceived as deceptive or insincere, often labeled as "stunts" or "gestures"

No Comments

No Known Activations