INDEX

Explanations

information related to criminal activities or investigations

oai_token-act-pair · gpt-3.5-turbo

negations related to arrests and injuries in a report

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

erity

-0.78

å§«

-0.75

gat

-0.74

uble

-0.74

Tes

-0.73

reddit

-0.70

DragonMagazine

-0.69

trap

-0.69

bey

-0.68

rypt

-0.68

POSITIVE LOGITS

 indication

1.11

 indications

1.06

 formal

0.96

 official

0.93

 kidding

0.91

 mention

0.89

 timetable

0.86

 conclusive

0.86

 signs

0.85

 definitive

0.84

Activations Density 0.127%

information related to criminal activities or investigations

negations related to arrests and injuries in a report

No Comments

No Known Activations