INDEX

Explanations

phrases that indicate specific structured plans or protocols

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

0:0.03

1:0.01

2:0.06

3:0.08

4:0.28

5:0.02

6:0.17

7:0.14

8:0.02

9:0.03

10:0.05

11:0.05

Negative Logits

elsen

-1.54

Aren

-1.49

afia

-1.39

osponsors

-1.36

eve

-1.35

Tonight

-1.35

ustomed

-1.33

◼

-1.31

abled

-1.28

orrow

-1.27

POSITIVE LOGITS

 basics

1.42

 Parables

1.38

 blank

1.31

emb

1.30

ittal

1.28

 incomplete

1.25

 pencil

1.25

 molten

1.23

:,

1.20

 Converted

1.18

Activations Density 0.021%

phrases that indicate specific structured plans or protocols

No Known Activations