INDEX

Explanations

terms related to control and authority

oai_token-act-pair · gpt-3.5-turbo

phrases related to power and dominance

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

à¤

-0.68

Recommend

-0.67

Bi

-0.65

TOM

-0.63

 Masquerade

-0.63

Intel

-0.62

é¾įå

-0.62

Matt

-0.61

Mich

-0.61

 Tale

-0.60

POSITIVE LOGITS

vier

0.92

orship

0.83

 control

0.82

iveness

0.80

ership

0.80

eering

0.78

 freak

0.77

taker

0.74

essee

0.74

rador

0.73

Activations Density 0.027%

terms related to control and authority

phrases related to power and dominance

No Comments

No Known Activations