INDEX

Explanations

visual elements and their attributes in the text

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L9_Hcat_z_lr1.20e-03_l11.20e+00_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.9.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.06

1:0.15

2:0.03

3:0.05

4:0.06

5:0.03

6:0.18

7:0.05

8:0.02

9:0.15

10:0.03

11:0.13

Negative Logits

 Matrix

-2.90

』

-2.75

 labou

-2.68

amphetamine

-2.63

 opium

-2.63

*)

-2.55

-2.54

 Camel

-2.53

 Hoffman

-2.51

ズ

-2.47

POSITIVE LOGITS

Courtesy

5.58

 Courtesy

4.86

 courtesy

4.76

tesy

4.44

Getty

3.28

ç

3.17

gerald

3.13

sonian

3.12

CHR

3.07

cour

3.06

Activations Density 0.001%

visual elements and their attributes in the text

No Known Activations