INDEX

Explanations

mentions of gruesome or violent incidents

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

words and phrases associated with various personal names and identities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-att-kk

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.3.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

0:0.02

1:0.03

2:0.09

3:0.27

4:0.02

5:0.02

6:0.13

7:0.09

8:0.04

9:0.08

10:0.07

11:0.09

Negative Logits

eworld

-1.18

pac

-1.17

PRES

-1.12

INT

-1.09

 Wonders

-1.09

 Alter

-1.07

venient

-1.04

alian

-1.03

fol

-1.03

aea

-1.03

POSITIVE LOGITS

�醒

1.24

�

1.23

Tube

1.23

ⓘ

1.22

aceae

1.22

 tumble

1.20

ornia

1.19

�士

1.08

ashtra

1.06

��

1.04

Activations Density 0.003%

mentions of gruesome or violent incidents

words and phrases associated with various personal names and identities

No Comments

No Known Activations