INDEX

Explanations

attends to tokens that denote numerical or structural elements from tokens indicating specific legal or procedural 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

0:0.07

1:0.06

2:0.08

3:0.11

4:0.06

5:0.04

6:0.44

7:0.10

Negative Logits

sizeCache

-0.45

BufferException

-0.38

queryInterface

-0.38

sidemargin

-0.36

interopRequire

-0.36

帖最后由

-0.36

Erreferentziak

-0.36

endregion

-0.36

Tembelea

-0.36

quelize

-0.35

POSITIVE LOGITS

\\

0.23

 Satisfied

0.23

 Galilee

0.23

win

0.22

 trai

0.22

 gắng

0.22

ziona

0.22

衅

0.22

ensi

0.22

 المعيارى

0.22

Activations Density 0.026%

attends to tokens that denote numerical or structural elements from tokens indicating specific legal or procedural terminology

No Comments

No Known Activations