INDEX

Explanations

attends to tokens marked by double asterisks that indicate a code-like structure from tokens marked by double square brackets that indicate context or closures

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

0:0.10

1:0.13

2:0.33

3:0.06

4:0.08

5:0.04

6:0.06

7:0.16

Negative Logits

 StatefulWidget

-0.30

NameInMap

-0.29

 StatelessWidget

-0.28

stdc

-0.25

textTheme

-0.24

ConstraintMaker

-0.24

RunAsync

-0.24

OrNil

-0.23

 unknownFields

-0.23

POSITIVE LOGITS

BibitemShut

0.38

 }}$}

0.35

\}}

0.35

RegressionTest

0.33

/−

0.32

\\

0.32

 combineReducers

0.31

 photolibrary

0.31

 насељу

0.30

 nyelven

0.29

Activations Density 0.408%

attends to tokens marked by double asterisks that indicate a code-like structure from tokens marked by double square brackets that indicate context or closures

No Comments

No Known Activations