INDEX

Explanations

ordinal numerical references, such as ranking words in a list based on position

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

New Auto-Interp

Top Features by Cosine Similarity

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.07

1:0.01

2:0.31

3:0.07

4:0.11

5:0.08

6:0.02

7:0.01

8:0.14

9:0.07

10:0.03

11:0.01

Negative Logits

 FIGHT

-1.26

tein

-1.26

ious

-1.20

iously

-1.18

Avg

-1.18

 Doodle

-1.16

Zin

-1.13

 Vance

-1.13

 behav

-1.11

LS

-1.11

POSITIVE LOGITS

mast

1.50

 horizon

1.45

aucas

1.41

ridges

1.36

士

1.36

end

1.33

 airports

1.32

 junction

1.31

 sidx

1.29

 mosqu

1.28

Activations Density 0.024%

ordinal numerical references, such as ranking words in a list based on position

No Comments

No Known Activations