INDEX

Explanations

occurrences of the word "in" related to various contexts

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

0:0.01

1:0.02

2:0.18

3:0.10

4:0.01

5:0.02

6:0.04

7:0.16

8:0.12

9:0.16

10:0.08

11:0.04

Negative Logits

ahs

-1.30

 shine

-1.29

 aside

-1.23

 greets

-1.21

rouse

-1.14

caster

-1.13

 counterpart

-1.11

nings

-1.11

ships

-1.11

 counterparts

-1.10

POSITIVE LOGITS

glomer

1.50

bably

1.49

retty

1.46

Pwr

1.37

iotic

1.28

nown

1.28

 TAMADRA

1.26

chwitz

1.26

��

1.25

 Codex

1.22

Activations Density 0.010%

occurrences of the word "in" related to various contexts

No Known Activations