INDEX

Explanations

mentions or references to someone named "Rom" with varying contexts

oai_token-act-pair · gpt-3.5-turbo

references to the city of Rome or related terms

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

insula

-0.76

gettable

-0.75

pressed

-0.68

 attendant

-0.68

IELD

-0.67

Ļ½

-0.67

 graded

-0.67

ctic

-0.63

 weights

-0.62

 eyed

-0.62

POSITIVE LOGITS

antics

1.04

ulus

1.03

anes

1.03

ijn

1.02

aji

1.01

atic

0.95

antic

0.94

antically

0.93

ano

0.92

ace

0.91

Activations Density 0.017%

mentions or references to someone named "Rom" with varying contexts

references to the city of Rome or related terms

No Comments

No Known Activations