INDEX

Explanations

attends to configuration-related tokens from earlier tokens in the sequence

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.13

1:0.20

2:0.16

3:0.12

4:0.16

5:0.04

6:0.07

7:0.07

Negative Logits

 للمعارف

-0.43

 "..\..\

-0.42

udesta

-0.40

AndEndTag

-0.38

ksesta

-0.34

ftagPool

-0.34

 EconPapers

-0.33

 initComponents

-0.32

prnewswire

-0.31

 GenerationType

-0.31

POSITIVE LOGITS

Chham

0.32

 استنادى

0.32

pédie

0.28

entera

0.27

idel

0.26

INTER

0.26

 Kuan

0.25

찌

0.25

 جغرافيا

0.25

hind

0.25

Activations Density 0.003%

attends to configuration-related tokens from earlier tokens in the sequence

No Comments

No Known Activations