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

Not in Any Lists

No Comments

Head Attr Weights

0:0.08

1:0.07

2:0.09

3:0.09

4:0.08

5:0.08

6:0.08

7:0.09

8:0.08

9:0.07

10:0.07

11:0.08

Negative Logits

 Punch

-2.05

 Wrap

-1.81

Roll

-1.78

 Drill

-1.73

Bench

-1.71

 punch

-1.68

Record

-1.64

 Oath

-1.63

 scalp

-1.63

Letter

-1.58

POSITIVE LOGITS

 civilisation

1.93

merce

1.89

ependence

1.72

 unification

1.59

NetMessage

1.58

adra

1.58

lda

1.57

ADRA

1.56

iless

1.55

 merging

1.55

Activations Density 0.000%

No Known Activations

This feature has no known activations.