INDEX

Explanations

phrases and references indicating strong attributes or quantities associated with concepts or objects

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

1:0.08

2:0.16

3:0.09

4:0.01

5:0.03

6:0.11

7:0.10

8:0.09

9:0.12

10:0.07

11:0.06

Negative Logits

 Hits

-1.41

buster

-1.39

 Picks

-1.17

gets

-1.11

rome

-1.09

Phones

-1.08

bats

-1.08

busters

-1.07

aves

-1.07

 Turns

-1.07

POSITIVE LOGITS

 overlap

1.34

 ambiguity

1.32

arial

1.31

 validity

1.29

 variance

1.28

 discrepancy

1.25

 shortage

1.24

 inconsistency

1.21

 similarity

1.20

 difference

1.19

Activations Density 0.125%

phrases and references indicating strong attributes or quantities associated with concepts or objects

No Comments

No Known Activations