INDEX

Explanations

references to physical locations where significant events or actions take place

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

phrases indicating positions of authority or power

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-att-kk

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.2.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.04

1:0.03

2:0.14

3:0.06

4:0.10

5:0.08

6:0.03

7:0.04

8:0.10

9:0.24

10:0.06

11:0.04

Negative Logits

ć

-1.47

д

-1.45

ь

-1.41

DragonMagazine

-1.39

zsche

-1.37

rencies

-1.32

iland

-1.28

maxwell

-1.27

gypt

-1.26

issance

-1.26

POSITIVE LOGITS

 seats

1.56

holder

1.48

holders

1.36

 seat

1.34

 holders

1.32

hung

1.31

 vacated

1.28

 threshold

1.26

arers

1.23

 skew

1.23

Activations Density 0.022%

references to physical locations where significant events or actions take place

phrases indicating positions of authority or power

No Comments

No Known Activations