INDEX

Explanations

attends to the token "new" from tokens associated with "with" in various contexts

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

0:0.31

1:0.24

2:0.08

3:0.05

4:0.05

5:0.06

6:0.04

7:0.12

Negative Logits

SequentialGroup

-0.34

DoubleQuotes

-0.34

/>";

-0.33

ampton

-0.31

principalTable

-0.30

//};

-0.29

.}(

-0.29

topper

-0.29

 NgModule

-0.29

ividual

-0.29

POSITIVE LOGITS

:✨

0.33

PYX

0.30

 beginnetje

0.29

таратура

0.28

orElseThrow

0.27

::::::::

0.27

OCCURRED

0.26

ResumeLayout

0.26

'{@

0.26

pagestyle

0.26

Activations Density 0.150%

attends to the token "new" from tokens associated with "with" in various contexts

No Comments

No Known Activations