INDEX

Explanations

phrases that indicate content from various sources, particularly those introducing or citing textual passages

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_L3_Hcat_z_lr1.20e-03_l19.00e-01_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.3.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

0:0.02

1:0.06

2:0.13

3:0.03

4:0.02

5:0.09

6:0.08

7:0.07

8:0.08

9:0.22

10:0.07

11:0.08

Negative Logits

aukee

-1.29

 swing

-1.27

 recovering

-1.27

 slump

-1.25

 headaches

-1.25

 coun

-1.20

 settle

-1.12

 reckoning

-1.12

 babys

-1.12

 productive

-1.12

POSITIVE LOGITS

 initials

1.43

 Dictionary

1.40

LU

1.36

 Koran

1.36

 insign

1.35

"@

1.35

 packets

1.35

 Emin

1.34

 acronym

1.31

solete

1.28

Activations Density 0.020%

phrases that indicate content from various sources, particularly those introducing or citing textual passages

No Comments

No Known Activations