INDEX

Explanations

product or brand names in capital letters

oai_token-act-pair · gpt-3.5-turbo

abbreviations and acronyms related to a particular domain or context

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

 Paul

-0.64

 Aman

-0.64

agra

-0.63

mary

-0.63

beg

-0.62

 Abraham

-0.62

 ammon

-0.61

 Jama

-0.61

 aisle

-0.61

 elect

-0.61

POSITIVE LOGITS

TL

4.61

tl

2.34

TL

2.18

TN

1.67

TF

1.27

TP

1.25

TG

1.23

TD

1.19

TPS

1.18

KT

1.17

Activations Density 0.010%

product or brand names in capital letters

abbreviations and acronyms related to a particular domain or context

No Comments

No Known Activations