INDEX

Explanations

attends to underscores separating token patterns that appear in the context of programming or coding terminology

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 10-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_10/width_16k/average_l0_70

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.10.attn.hook_z

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.17

1:0.20

2:0.17

3:0.07

4:0.06

5:0.06

6:0.08

7:0.15

Negative Logits

جغرافيا

-0.32

__(/*!

-0.30

 pierdas

-0.30

ailangan

-0.29

ือก

-0.29

 Biôgrafia

-0.29

GTCX

-0.29

 pinulongan

-0.28

исленность

-0.28

pearing

-0.28

POSITIVE LOGITS

 فريبيس

0.30

 propOrder

0.28

UDC

0.28

 móvel

0.28

gameserver

0.26

mappedBy

0.26

]<<"

0.24

auc

0.23

izien

0.23

 Plato

0.23

Activations Density 0.679%

attends to underscores separating token patterns that appear in the context of programming or coding terminology

No Comments

No Known Activations