INDEX

Explanations

formal titles and honorifics

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.02

1:0.03

2:0.07

3:0.03

4:0.04

5:0.04

6:0.52

7:0.04

8:0.03

9:0.03

10:0.07

11:0.03

Negative Logits

 Morse

-1.43

 destro

-1.41

uador

-1.39

 outweigh

-1.35

CFR

-1.34

�

-1.34

\\\\\\\\

-1.31

anyahu

-1.31

 persecution

-1.31

 outwe

-1.31

POSITIVE LOGITS

xus

1.76

oslav

1.57

ρ

1.51

hid

1.38

igon

1.37

opher

1.35

mitter

1.34

identified

1.34

ulhu

1.34

idate

1.33

Activations Density 0.042%

formal titles and honorifics

No Known Activations