INDEX

Explanations

attends to various items and features marked as "co" from tokens later 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

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.09

1:0.11

2:0.11

3:0.13

4:0.11

5:0.08

6:0.16

7:0.16

Negative Logits

IVEREF

-0.37

 calendriers

-0.36

 مشين

-0.36

횟

-0.35

FormTagHelper

-0.35

WebMethod

-0.34

QMetaType

-0.34

keyColumn

-0.32

RenderAtEndOf

-0.32

SharedCtor

-0.31

POSITIVE LOGITS

jsonwebtoken

0.27

 Mumford

0.27

OSD

0.27

mybatisplus

0.26

CodeAnalysis

0.26

VersionUID

0.25

 olympiques

0.25

плек

0.24

filepath

0.24

MultipartFile

0.24

Activations Density 0.000%

attends to various items and features marked as "co" from tokens later in the sequence

No Comments

No Known Activations