INDEX

Explanations

terms related to biological and medical concepts

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

0:0.08

1:0.18

2:0.02

3:0.03

4:0.02

5:0.32

6:0.04

7:0.02

8:0.05

9:0.10

10:0.08

11:0.02

Negative Logits

 LIMITED

-1.93

theless

-1.72

 CONS

-1.64

 communications

-1.62

terday

-1.58

 Simon

-1.58

 colle

-1.57

 ASUS

-1.57

DEL

-1.57

NYC

-1.54

POSITIVE LOGITS

onga

2.46

bug

2.34

rage

2.22

adish

2.17

roe

2.12

term

2.11

exc

2.10

eral

2.08

inis

2.07

oqu

2.06

Activations Density 0.143%

terms related to biological and medical concepts

No Known Activations