INDEX

Explanations

names, likely focusing on first names

oai_token-act-pair · gpt-3.5-turbo

names of people, particularly those involved in notable events or 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

iance

-0.89

itect

-0.88

aic

-0.87

ibel

-0.77

iers

-0.77

inar

-0.74

yrinth

-0.74

oÄŁ

-0.74

efficient

-0.73

ectar

-0.72

POSITIVE LOGITS

 Rogers

1.02

 Bobby

0.99

 Doodle

0.89

 Willie

0.84

 Benny

0.81

 Burns

0.79

 Pett

0.79

 Tyson

0.79

 Horton

0.79

 Hank

0.78

Activations Density 0.047%

names, likely focusing on first names

names of people, particularly those involved in notable events or contexts

No Comments

No Known Activations