INDEX

Explanations

phrases related to surprising or unknowingly discovered information

oai_token-act-pair · gpt-3.5-turbo

phrases that express awareness or knowledge

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

ĪĴ

-0.85

assi

-0.78

stros

-0.76

utic

-0.75

uckles

-0.75

ŃĶ

-0.74

thren

-0.74

ressor

-0.73

empl

-0.70

elin

-0.70

POSITIVE LOGITS

 anymore

0.92

yet

0.80

 whatsoever

0.77

DERR

0.75

ledge

0.74

 beforehand

0.74

 wrongdoing

0.73

nor

0.72

 anything

0.72

 aloud

0.71

Activations Density 0.092%

phrases related to surprising or unknowingly discovered information

phrases that express awareness or knowledge

No Comments

No Known Activations