INDEX

Explanations

references to televised news shows and interviews

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_L9_Hcat_z_lr1.20e-03_l11.20e+00_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.9.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

0:0.13

1:0.02

2:0.03

3:0.32

4:0.02

5:0.15

6:0.02

7:0.04

8:0.09

9:0.01

10:0.10

11:0.02

Negative Logits

 shielding

-2.00

 transl

-1.93

yrights

-1.91

 deposits

-1.90

 markings

-1.90

umb

-1.85

 localization

-1.84

 labou

-1.84

 transfers

-1.83

 dispers

-1.82

POSITIVE LOGITS

 Oprah

2.61

 Tonight

2.38

Talk

2.33

 Hannity

2.23

 CNBC

2.15

 Mysteries

2.08

 Breakfast

2.05

 Rush

2.01

Week

2.01

 Talk

2.00

Activations Density 0.334%

references to televised news shows and interviews

No Comments

No Known Activations