INDEX

Explanations

names of individuals, particularly in contexts involving authority or decisions

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

New Auto-Interp

Top Features by Cosine Similarity

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

0:0.02

1:0.16

2:0.08

3:0.05

4:0.01

5:0.03

6:0.11

7:0.10

8:0.11

9:0.04

10:0.11

11:0.11

Negative Logits

 deduction

-1.24

..."

-1.07

 Grateful

-1.05

acebook

-1.03

 concess

-1.03

 blackmail

-1.03

 deductions

-1.03

 stadiums

-1.02

 deduct

-1.02

 disapprove

-1.01

POSITIVE LOGITS

inski

1.30

覚醒

1.19

onen

1.19

opal

1.15

�

1.15

Sym

1.14

stein

1.12

opoulos

1.12

insk

1.11

ason

1.10

Activations Density 0.117%

names of individuals, particularly in contexts involving authority or decisions

No Comments

No Known Activations