INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L9_Hcat_z_lr1.20e-03_l11.20e+00_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.9.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

0:0.10

1:0.08

2:0.08

3:0.09

4:0.07

5:0.06

6:0.07

7:0.07

8:0.09

9:0.08

10:0.07

11:0.08

Negative Logits

 Disco

-3.06

 techno

-2.85

 Capitalism

-2.84

Fn

-2.82

 Thou

-2.78

 Leviathan

-2.78

 utopian

-2.77

 Zionism

-2.76

 Parad

-2.75

 Lite

-2.73

POSITIVE LOGITS

icut

3.16

Secretary

2.91

 inspector

2.79

FBI

2.76

ridge

2.71

 scrub

2.68

 counselor

2.66

Department

2.65

orneys

2.65

onnaissance

2.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.