INDEX

Explanations

attends to unique tokens found in the middle of the sequences from the far right tokens that are often related to names or specific identifiers

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

0:0.28

1:0.26

2:0.13

3:0.05

4:0.05

5:0.02

6:0.07

7:0.10

Negative Logits

 myſelf

-0.32

 Erscheinung

-0.31

 antaranya

-0.30

 Shakspeare

-0.30

 francès

-0.29

 occaf

-0.29

 picioare

-0.29

 raiſ

-0.29

 themſelves

-0.28

 Houſe

-0.28

POSITIVE LOGITS

providedIn

0.38

ⓧ

0.38

+#+#

0.37

 Réponses

0.34

TemporalType

0.34

XmlAccessType

0.34

ArrowToggle

0.33

httphttps

0.33

ंदीखरीदारी

0.33

HtmlAttribute

0.31

Activations Density 0.396%

attends to unique tokens found in the middle of the sequences from the far right tokens that are often related to names or specific identifiers

No Comments

No Known Activations