INDEX

Explanations

evaluative language reflecting quality or effectiveness

New Auto-Interp

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

0:0.01

1:0.02

2:0.07

3:0.08

4:0.16

5:0.02

6:0.38

7:0.09

8:0.02

9:0.02

10:0.04

11:0.04

Negative Logits

 Bots

-1.32

 cassette

-1.25

 slot

-1.25

 parallels

-1.20

 logo

-1.18

 suffix

-1.18

 Bloody

-1.14

 Equal

-1.14

 isot

-1.11

 Binding

-1.10

POSITIVE LOGITS

ouver

1.73

aign

1.51

ensibly

1.50

aspers

1.49

abit

1.49

lished

1.48

yout

1.48

urious

1.46

arters

1.45

aughs

1.36

Activations Density 0.025%

evaluative language reflecting quality or effectiveness

No Known Activations