INDEX

Explanations

attends to certain identifier tokens from specific examples or data points within explanations

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

0:0.06

1:0.07

2:0.06

3:0.09

4:0.08

5:0.05

6:0.43

7:0.12

Negative Logits

 iNdEx

-0.31

zzar

-0.30

SizeMode

-0.30

 strangely

-0.30

 kasarigan

-0.29

 apart

-0.29

 TextAppearance

-0.28

 Peña

-0.28

ölker

-0.28

Morfologia

-0.27

POSITIVE LOGITS

0.37



0.35

}")

0.34

}.

0.33

 périodes

0.32

]`

0.32

""}

0.31

 cref

0.31

"));

0.31

istoitu

0.30

Activations Density 1.748%

attends to certain identifier tokens from specific examples or data points within explanations

No Comments

No Known Activations