INDEX

Explanations

controversial or attention-grabbing content

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

words that indicate moral judgment or critique

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

0:0.08

1:0.04

2:0.10

3:0.08

4:0.05

5:0.11

6:0.04

7:0.02

8:0.19

9:0.13

10:0.06

11:0.03

Negative Logits

 belie

-1.34

enment

-1.33

SHARE

-1.24

awaru

-1.16

 Ethan

-1.15

 airst

-1.14

regon

-1.12

 hemor

-1.10

fly

-1.09

(£

-1.09

POSITIVE LOGITS

put

1.42

added

1.34

reated

1.31

ired

1.30

APP

1.28

active

1.26

ind

1.26

pir

1.25

oxic

1.24

acceptable

1.19

Activations Density 0.063%

controversial or attention-grabbing content

words that indicate moral judgment or critique

No Comments

No Known Activations