INDEX

Explanations

conjunctions and phrases linked to physical connections or relationships

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

0:0.04

1:0.01

2:0.05

3:0.06

4:0.04

5:0.04

6:0.45

7:0.05

8:0.04

9:0.04

10:0.09

11:0.06

Negative Logits

autions

-1.80

ilial

-1.70

 Pradesh

-1.62

earances

-1.49

 NETWORK

-1.35

umatic

-1.34

neau

-1.29

itions

-1.29

 installs

-1.28

ionage

-1.27

POSITIVE LOGITS

��

1.40

�

1.39

waters

1.30

TextColor

1.27

Loading

1.19

 Queens

1.19

onen

1.18

obin

1.18

 widow

1.15

Machine

1.14

Activations Density 0.000%

conjunctions and phrases linked to physical connections or relationships

No Known Activations