INDEX

Explanations

the letter "r"

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

Head Attr Weights

0:0.08

1:0.08

2:0.08

3:0.09

4:0.07

5:0.06

6:0.08

7:0.07

8:0.07

9:0.09

10:0.09

11:0.08

Negative Logits

Sov

-2.85

Bet

-2.82

 Ambro

-2.80

 Monaco

-2.68

akov

-2.66

aco

-2.65

oan

-2.65

 Tomato

-2.65

 Medieval

-2.64

agall

-2.63

POSITIVE LOGITS

isodes

2.79

taboola

2.63

 aims

2.59

 Lindsay

2.55

 referrals

2.51

 diagnoses

2.51

 focuses

2.47

 target

2.43

HR

2.40

 resid

2.39

Activations Density 0.000%

the letter "r"

No Known Activations