INDEX

Explanations

names of individuals

oai_token-act-pair · gpt-3.5-turbo

names of individuals associated with key political figures

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 ellipt

-0.90

 answ

-0.80

 mathemat

-0.75

 deflation

-0.69

osc

-0.68

orter

-0.67

 dilig

-0.67

 magistrate

-0.67

ort

-0.66

 laund

-0.66

POSITIVE LOGITS

 Haley

3.04

Fay

1.71

 Vance

1.35

 Cummings

1.33

Mia

1.25

Nem

1.18

 Frazier

1.16

 Kislyak

1.11

 Mesa

1.09

 Cindy

1.08

Activations Density 0.043%

names of individuals

names of individuals associated with key political figures

No Comments

No Known Activations