INDEX

Explanations

words related to rules, requests, and instructions

oai_token-act-pair · gpt-3.5-turbo

references to rules or guidelines

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

hung

-0.73

heid

-0.73

joice

-0.67

ãĢĲ

-0.66

jet

-0.65

joy

-0.65

 Fever

-0.63

 Doctors

-0.63

vict

-0.62

kai

-0.62

POSITIVE LOGITS

UL

1.10

ANE

1.05

tymology

0.96

ULE

0.94

OAD

0.93

ATING

0.93

ATION

0.93

OUS

0.91

VIDIA

0.91

NER

0.91

Activations Density 0.013%

words related to rules, requests, and instructions

references to rules or guidelines

No Comments

No Known Activations