INDEX

Explanations

prepositions and phrases indicating relationships or actions performed by subjects

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

1:0.01

2:0.22

3:0.05

4:0.14

5:0.04

6:0.03

7:0.18

8:0.05

9:0.04

10:0.07

11:0.09

Negative Logits

ankind

-2.16

abet

-2.15

ividual

-2.05

selves

-1.84

ergic

-1.75

bda

-1.73

orer

-1.63

sensitive

-1.61

Surv

-1.61

 heir

-1.61

POSITIVE LOGITS

��極

2.19

 customary

2.03

��

1.77

ワ

1.75

 doubles

1.57

 sticker

1.56

 unbeat

1.56

 motto

1.53

 disqual

1.53

 draconian

1.53

Activations Density 0.001%

prepositions and phrases indicating relationships or actions performed by subjects

No Known Activations