INDEX

Explanations

phrases that emphasize questioning or dialogue

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_L9_Hcat_z_lr1.20e-03_l11.20e+00_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.9.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

0:0.11

1:0.02

2:0.05

3:0.05

4:0.05

5:0.06

6:0.19

7:0.05

8:0.04

9:0.27

10:0.03

11:0.03

Negative Logits

Bo

-3.56

 Alph

-3.42

Balt

-3.29

 looted

-3.26

 looting

-3.25

 armour

-3.24

 Baal

-3.18

Account

-3.18

 Balt

-3.18

ails

-3.08

POSITIVE LOGITS

 Wendy

8.74

 Wend

5.02

wid

4.17

 Randall

4.15

kefeller

4.03

Rand

3.88

WF

3.85

 Perkins

3.68

FW

3.64

Syd

3.62

Activations Density 0.000%

phrases that emphasize questioning or dialogue

No Comments

No Known Activations