INDEX

Explanations

architecture-related terms, specifically related to floors

oai_token-act-pair · gpt-3.5-turbo

references to the floor in various contexts

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

eday

-0.81

enegger

-0.75

 Unch

-0.69

ortium

-0.68

ALS

-0.67

agnar

-0.67

elson

-0.67

eness

-0.66

acca

-0.64

 Alonso

-0.64

POSITIVE LOGITS

floor

1.11

boards

1.02

 floor

1.02

 Floor

0.97

nuts

0.95

 floors

0.94

 mats

0.93

plan

0.93

parts

0.89

stairs

0.85

Activations Density 0.013%

architecture-related terms, specifically related to floors

references to the floor in various contexts

No Comments

No Known Activations