INDEX

Explanations

numerical values and their associated context

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

Head Attr Weights

0:0.02

1:0.02

2:0.19

3:0.11

4:0.06

5:0.04

6:0.13

7:0.03

8:0.04

9:0.06

10:0.07

11:0.18

Negative Logits

afort

-1.68

iliate

-1.67

 recons

-1.54

ilaterally

-1.51

aples

-1.50

 thereof

-1.49

hesda

-1.48

 Donation

-1.48

govtrack

-1.48

 theirs

-1.44

POSITIVE LOGITS

dB

1.43

�

1.40

Cla

1.40

ハ

1.39

tons

1.35

 Aust

1.33

 Percent

1.33

 loud

1.33

 whatsoever

1.32

aws

1.32

Activations Density 0.018%

numerical values and their associated context

No Known Activations