INDEX

Explanations

phrases related to revealing hidden or important information

oai_token-act-pair · gpt-3.5-turbo

phrases relating to hidden dangers or underlying issues

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

Merit

-0.87

ailability

-0.78

ãĥīãĥ©ãĤ´ãĥ³

-0.77

FACE

-0.74

 Klux

-0.67

 divest

-0.67

SPONSORED

-0.66

owship

-0.65

çĦ

-0.65

 effic

-0.64

POSITIVE LOGITS

 iceberg

0.87

yip

0.83

 Racer

0.70

ppy

0.67

Rai

0.67

 Direction

0.65

ora

0.64

chio

0.63

ariat

0.63

ogly

0.63

Activations Density 0.149%

phrases related to revealing hidden or important information

phrases relating to hidden dangers or underlying issues

No Comments

No Known Activations