INDEX

Explanations

phrases related to entertainment and media, such as TV show titles and plot elements

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

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

Not in Any Lists

No Comments

Head Attr Weights

0:0.06

1:0.02

2:0.09

3:0.09

4:0.03

5:0.08

6:0.02

7:0.05

8:0.24

9:0.10

10:0.11

11:0.06

Negative Logits

ablishment

-1.18

 Employ

-1.03

 jurisd

-1.02

��

-1.02

 Attend

-1.01

ensable

-0.99

AST

-0.99

ertation

-0.97

yne

-0.97

istence

-0.95

POSITIVE LOGITS

¢

1.16

 volleyball

1.08

 commercials

1.07

 puzzles

1.03

 drums

1.03

 games

1.02

 keyboards

1.01

 dram

1.01

 poker

1.00

bol

0.98

Activations Density 0.115%

phrases related to entertainment and media, such as TV show titles and plot elements

No Comments

No Known Activations