INDEX

Explanations

words and phrases indicating actions or processes related to bypassing or circumventing something

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

New Auto-Interp

Top Features by Cosine Similarity

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

0:0.05

1:0.02

2:0.21

3:0.04

4:0.28

5:0.04

6:0.02

7:0.03

8:0.11

9:0.08

10:0.05

11:0.02

Negative Logits

ufact

-1.62

isf

-1.44

aughters

-1.39

birds

-1.38

atures

-1.28

 cryst

-1.28

akeru

-1.28

cill

-1.25

 embr

-1.24

 advoc

-1.24

POSITIVE LOGITS

 charms

1.41

atche

1.36

 charm

1.33

ּ

1.24

Rect

1.21

 Bastard

1.19

Thumbnail

1.18

 Hole

1.17

 mistaken

1.17

 Vaughn

1.16

Activations Density 0.002%

words and phrases indicating actions or processes related to bypassing or circumventing something

No Comments

No Known Activations