INDEX

Explanations

references to firing or shooting incidents

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "fired" and its variations, indicating actions related to discharging weapons

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

Label

-0.70

Virgin

-0.69

DonaldTrump

-0.68

é¾įå

-0.68

é¾įå¥ĳå£«

-0.66

 Obesity

-0.66

este

-0.65

 amen

-0.65

 Proud

-0.65

Syd

-0.64

POSITIVE LOGITS

balls

0.93

flies

0.86

fights

0.86

 gunshots

0.85

 volley

0.84

FIELD

0.83

bomb

0.82

 firing

0.81

 gunfire

0.80

 artillery

0.80

Activations Density 0.027%

references to firing or shooting incidents

occurrences of the word "fired" and its variations, indicating actions related to discharging weapons

No Comments

No Known Activations