INDEX

Explanations

references to demographic groups and their respective nationalities

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L10_Hcat_z_lr1.20e-03_l11.30e+00_ds24576_bs4096_dc1.00e-05_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.10.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.05

1:0.03

2:0.02

3:0.14

4:0.14

5:0.02

6:0.04

7:0.19

8:0.02

9:0.05

10:0.04

11:0.21

Negative Logits

 Italy

-2.11

Italy

-2.07

 scapego

-2.06

 adultery

-2.02

epend

-1.98

 Countries

-1.95

 blame

-1.91

elfare

-1.86

cia

-1.85

ornia

-1.83

POSITIVE LOGITS

 Jordanian

2.41

Sund

2.40

Indian

2.35

 Pakistani

2.28

 Egyptian

2.17

'."

2.14

 Chilean

2.13

 Venezuelan

2.09

 Korean

2.08

 Norwegian

2.04

Activations Density 0.026%

references to demographic groups and their respective nationalities

No Known Activations