INDEX

Explanations

either names/words related to specific individuals, locations, or terms, or words related to conflicts or atrocities

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

elements related to specific names or titles

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

0:0.10

1:0.02

2:0.24

3:0.07

4:0.16

5:0.07

6:0.03

7:0.02

8:0.06

9:0.10

10:0.05

11:0.02

Negative Logits

 Rove

-1.23

SPONSORED

-1.18

avorite

-1.17

 concess

-1.15

 defe

-1.12

pse

-1.11

 Samoa

-1.09

-1.07

acebook

-1.06

VERTISEMENT

-1.04

POSITIVE LOGITS

aja

1.46

vous

1.34

qi

1.27

 Thrust

1.24

agus

1.24

ut

1.23

ondo

1.22

️

1.21

hin

1.20

atha

1.20

Activations Density 0.002%

either names/words related to specific individuals, locations, or terms, or words related to conflicts or atrocities

elements related to specific names or titles

No Comments

No Known Activations