INDEX

Explanations

phrases related to programming, code control, and technical instructions

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

terms related to control and authority

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-att-kk

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.3.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.03

1:0.02

2:0.11

3:0.07

4:0.02

5:0.04

6:0.10

7:0.30

8:0.07

9:0.03

10:0.07

11:0.08

Negative Logits

 Schwar

-1.22

 Ruin

-1.22

DERR

-1.19

 Prosper

-1.15

Nir

-1.14

BuyableInstoreAndOnline

-1.09

 misunder

-1.07

ospons

-1.03

 Defeat

-1.01

 Wonderful

-1.00

POSITIVE LOGITS

ament

1.15

EH

1.09

measures

1.04

engers

1.03

 controls

1.02

semble

1.02

EF

1.02

 circuitry

1.01

eg

1.00

 detects

1.00

Activations Density 0.035%

phrases related to programming, code control, and technical instructions

terms related to control and authority

No Comments

No Known Activations