INDEX

Explanations

instances where someone is sitting down at a table with another person

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

0:0.06

1:0.02

2:0.24

3:0.08

4:0.13

5:0.04

6:0.04

7:0.02

8:0.14

9:0.10

10:0.05

11:0.03

Negative Logits

wallet

-1.29

said

-1.23

ache

-1.21

Minecraft

-1.06

REDACTED

-1.06

ILA

-1.06

doms

-1.05

iston

-1.03

amins

-1.03

govtrack

-1.03

POSITIVE LOGITS

ibaba

1.31

opol

1.19

pmwiki

1.18

�

1.09

onial

1.09

eper

1.08

lycer

1.06

eenth

1.06

Ops

1.05

 lookout

1.05

Activations Density 0.002%

instances where someone is sitting down at a table with another person

No Known Activations