INDEX

Explanations

references to criminal activities, specifically violent acts

oai_token-act-pair · gpt-3.5-turbo

instances of violent events

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

irez

-0.82

alde

-0.80

parser

-0.73

vik

-0.72

UTF

-0.72

 GOODMAN

-0.71

oen

-0.68

Ģ

-0.68

abad

-0.68

anto

-0.68

POSITIVE LOGITS

 Commodore

0.77

 Lovecraft

0.76

 Horus

0.73

plane

0.71

 Paran

0.70

 trumpet

0.68

 Starship

0.65

 Grateful

0.65

 Cadillac

0.65

 Buckley

0.64

Activations Density 0.000%

references to criminal activities, specifically violent acts

instances of violent events

No Comments

No Known Activations