INDEX

Explanations

politically charged phrases and names

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.13

1:0.03

2:0.08

3:0.12

4:0.03

5:0.09

6:0.04

7:0.02

8:0.25

9:0.07

10:0.07

11:0.03

Negative Logits

 Cord

-1.22

 Colossus

-1.09

tnc

-1.09

 Legacy

-1.05

 Doors

-1.05

 Views

-1.03

 Expedition

-1.02

senal

-1.01

-1.00

 Kashmir

-0.99

POSITIVE LOGITS

 anymore

1.60

yet

1.40

nor

1.35

 anything

1.18

anything

1.16

unless

1.15

yet

1.14

necess

1.12

nor

1.12

bud

1.12

Activations Density 0.213%

politically charged phrases and names

No Comments

No Known Activations