INDEX

Explanations

power-related words and terms, such as "powers," "authority," "exercise," and "discretion."

oai_token-act-pair · gpt-3.5-turbo

terms associated with governmental or legal authority and powers

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

adish

-0.80

 Despair

-0.73

 Temper

-0.73

 Tiff

-0.71

 Scarlet

-0.71

 Stain

-0.70

 Waves

-0.69

 Wheat

-0.69

hesis

-0.68

 Murd

-0.68

POSITIVE LOGITS

 vested

1.17

 delegated

1.15

 granted

1.02

 conferred

1.01

 afforded

0.85

 entrusted

0.84

 entitle

0.82

 constitutionally

0.80

 unchecked

0.79

 discretionary

0.79

Activations Density 0.098%

power-related words and terms, such as "powers," "authority," "exercise," and "discretion."

terms associated with governmental or legal authority and powers

No Comments

No Known Activations