INDEX

Explanations

the word "to"

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

1:0.08

2:0.08

3:0.08

4:0.07

5:0.08

6:0.07

7:0.08

8:0.11

9:0.07

10:0.07

11:0.08

Negative Logits

Sher

-1.85

 Lite

-1.82

 Feldman

-1.67

Sr

-1.63

 Colbert

-1.62

Kyr

-1.61

 backstory

-1.58

Detailed

-1.57

 Sherlock

-1.55

 scaff

-1.54

POSITIVE LOGITS

2.12

irable

2.06

 =====

2.02

licts

1.94

ilan

1.94

benefit

1.76

affected

1.76

lict

1.73

Reviewed

1.69

 ACTIONS

1.68

Activations Density 0.000%

the word "to"

No Known Activations