INDEX

Explanations

No Explanations Found

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

0:0.07

1:0.04

2:0.09

3:0.08

4:0.08

5:0.09

6:0.07

7:0.10

8:0.08

9:0.08

10:0.09

11:0.08

Negative Logits

ecause

-1.70

 wiret

-1.70

oop

-1.69

 Cheong

-1.59

aturdays

-1.55

◼

-1.53

hift

-1.51

 Directions

-1.48

 murd

-1.48

rosc

-1.46

POSITIVE LOGITS

arity

1.58

represented

1.53

 alienated

1.48

ilde

1.48

 nont

1.40

 compl

1.38

 plaintiffs

1.36

 excluded

1.36

 ambitious

1.34

 heterogeneity

1.34

Activations Density 0.000%

No Known Activations

This feature has no known activations.