INDEX

Explanations

expressions encouraging action or opinion-related responses

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_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.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

0:0.01

1:0.01

2:0.09

3:0.15

4:0.15

5:0.02

6:0.03

7:0.30

8:0.02

9:0.03

10:0.08

11:0.05

Negative Logits

 fixtures

-1.32

killers

-1.28

 pandemonium

-1.24

 fixture

-1.22

igate

-1.20

aneously

-1.17

iaz

-1.15

 isEnabled

-1.14

 overnight

-1.14

 yuan

-1.14

POSITIVE LOGITS

wise

1.35

 Paradox

1.32

 🙂

1.31

 Subtle

1.30

 cynicism

1.28

 readers

1.28

 anecdotal

1.26

 knowing

1.24

 wise

1.23

 Brave

1.23

Activations Density 0.253%

expressions encouraging action or opinion-related responses

No Comments

No Known Activations