INDEX

Explanations

phrases indicating the presence of a written or published work

oai_token-act-pair · gpt-3.5-turbo

references to written works or articles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

elsius

-0.82

 Predators

-0.78

 Answer

-0.66

 Nadu

-0.61

 Predator

-0.61

 Snapdragon

-0.59

edient

-0.58

elight

-0.58

 princes

-0.58

 Complex

-0.58

POSITIVE LOGITS

meal

1.79

toe

0.82

 piece

0.82

glass

0.80

work

0.78

fare

0.78

book

0.77

umen

0.76

hook

0.74

age

0.73

Activations Density 0.015%

phrases indicating the presence of a written or published work

references to written works or articles

No Comments

No Known Activations