INDEX

Explanations

people's names, specifically the name "Andre"

oai_token-act-pair · gpt-3.5-turbo

mentions of the name "Andre" 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 @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

pc

-0.72

 disp

-0.71

NZ

-0.70

 chambers

-0.70

nda

-0.68

 sequest

-0.68

=>

-0.67

 Flake

-0.67

SC

-0.66

psc

-0.65

POSITIVE LOGITS

 Andre

3.44

Andre

3.05

andre

1.55

 Antonio

1.43

 Pavel

1.37

 Mikhail

1.37

 Nikol

1.35

 Vlad

1.33

 Serge

1.26

 Sergei

1.23

Activations Density 0.014%

people's names, specifically the name "Andre"

mentions of the name "Andre" in various contexts

No Comments

No Known Activations