INDEX

Explanations

names of historical figures and locations

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

0:0.07

1:0.03

2:0.33

3:0.07

4:0.14

5:0.04

6:0.02

7:0.03

8:0.04

9:0.12

10:0.04

11:0.02

Negative Logits

taboola

-1.72

 Tradable

-1.45

 subp

-1.41

ICLE

-1.39

 Tsukuyomi

-1.34

��

-1.33

BILITIES

-1.31

���

-1.28

inational

-1.25

ibilities

-1.25

POSITIVE LOGITS

gow

1.40

iday

1.34

beck

1.26

essee

1.16

bender

1.13

 confid

1.12

erella

1.10

igans

1.09

 hangs

1.09

bart

1.08

Activations Density 0.027%

names of historical figures and locations

No Known Activations