INDEX

Explanations

common terms and phrases related to programming and technical concepts

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.2.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

1:0.03

2:0.03

3:0.07

4:0.03

5:0.05

6:0.11

7:0.11

8:0.04

9:0.06

10:0.23

11:0.11

Negative Logits

enegger

-0.91

ONSORED

-0.87

)."

-0.81

instead

-0.81

)</

-0.79

inki

-0.75

swick

-0.75

sic

-0.74

'."

-0.74

Rated

-0.69

POSITIVE LOGITS

Rat

0.67

 Phar

0.67

 rhetorical

0.65

\":

0.65

 Marlins

0.65

 concede

0.62

 Hatt

0.61

 Chomsky

0.61

 Roose

0.61

 Goodell

0.59

Activations Density 1.531%

common terms and phrases related to programming and technical concepts

No Known Activations