INDEX

Explanations

phrases related to criticism and legal investigations

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

words indicating responsibility or blame in a narrative context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

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

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_post

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

asse

-0.76

DragonMagazine

-0.72

glers

-0.72

 ILCS

-0.70

edin

-0.69

emption

-0.68

eret

-0.68

mint

-0.68

idem

-0.68

event

-0.67

POSITIVE LOGITS

 vowed

1.29

 likened

1.26

 praised

1.18

 urged

1.17

 denounced

1.15

 accused

1.14

 blamed

1.14

 lamented

1.13

 warned

1.13

 demanded

1.11

Activations Density 0.312%

phrases related to criticism and legal investigations

words indicating responsibility or blame in a narrative context

No Comments

No Known Activations