INDEX

Explanations

dates and events mentioned in news articles

oai_token-act-pair · gpt-3.5-turbo

Triggered by @bot

numerical representations of time or duration

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-att-kk

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

2:0.17

3:0.11

4:0.07

5:0.11

6:0.05

7:0.02

8:0.06

9:0.11

10:0.08

11:0.04

Negative Logits

 Corpus

-1.23

 Sagan

-1.12

 Lerner

-1.10

 exagger

-1.06

Ni

-1.04

 Organ

-1.01

 Sachs

-1.01

 Qatar

-1.01

ORN

-1.00

Cell

-1.00

POSITIVE LOGITS

umatic

1.57

heastern

1.44

achev

1.33

undred

1.29

axy

1.29

sembly

1.26

 respectively

1.26

illion

1.22

ittal

1.21

isode

1.20

Activations Density 0.009%

dates and events mentioned in news articles

numerical representations of time or duration

No Known Activations