INDEX

Explanations

personal descriptions or attributes assigned to individuals

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.3.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.02

1:0.05

2:0.06

3:0.04

4:0.02

5:0.04

6:0.05

7:0.07

8:0.08

9:0.30

10:0.09

11:0.11

Negative Logits

irtual

-1.41

thank

-1.36

endum

-1.30

NOTE

-1.30

onne

-1.28

rote

-1.28

abases

-1.25

hooting

-1.23

ickets

-1.18

obbies

-1.18

POSITIVE LOGITS

 cannibal

1.41

 indifferent

1.39

 unfit

1.38

 unworthy

1.36

 unfairly

1.33

��

1.30

 casts

1.30

 unethical

1.29

 subversive

1.29

 immoral

1.27

Activations Density 0.008%

personal descriptions or attributes assigned to individuals

No Known Activations