INDEX

Explanations

mentions of people's names or locations

oai_token-act-pair · gpt-3.5-turbo

terms related to academic and institutional authority

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

Mub

-0.69

isSpecialOrderable

-0.67

 showc

-0.65

hur

-0.65

pload

-0.65

ccoli

-0.63

 Bever

-0.62

iless

-0.62

wcs

-0.62

mercial

-0.60

POSITIVE LOGITS

ipel

0.80

ALLY

0.78

RECT

0.77

 omission

0.75

×¨

0.72

itely

0.71

ional

0.69

ANCE

0.68

×Ļ

0.64

ajor

0.64

Activations Density 0.287%

mentions of people's names or locations

terms related to academic and institutional authority

No Comments

No Known Activations