INDEX

Explanations

words related to properties or processes

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "pro" or professionalism

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

 needles

-0.74

gow

-0.71

 spears

-0.70

 Twain

-0.70

 Cornell

-0.68

 Dwell

-0.67

 Halls

-0.67

 woodland

-0.67

 Graveyard

-0.66

bodied

-0.66

POSITIVE LOGITS

digy

1.25

blems

1.15

secut

1.09

posal

1.06

secution

1.03

gressive

1.01

ceed

0.98

gression

0.96

hibited

0.93

verbs

0.93

Activations Density 0.015%

words related to properties or processes

references to the concept of "pro" or professionalism

No Comments

No Known Activations