INDEX

Explanations

phrase structures related to splitting or dividing

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.2.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

1:0.02

2:0.07

3:0.06

4:0.39

5:0.04

6:0.02

7:0.02

8:0.13

9:0.08

10:0.05

11:0.01

Negative Logits

tti

-1.63

Fey

-1.39

�

-1.35

audi

-1.34

jan

-1.28

mund

-1.27

ii

-1.26

adv

-1.25

na

-1.24

vis

-1.21

POSITIVE LOGITS

 evenly

1.60

 halves

1.60

 horizontally

1.46

imilar

1.39

otaur

1.38

 membrane

1.37

ormonal

1.36

apeshifter

1.35

oxide

1.33

 dogma

1.32

Activations Density 0.005%

phrase structures related to splitting or dividing

No Known Activations