INDEX

Explanations

attends to the token "label" from tokens of various markers

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

0:0.08

1:0.09

2:0.12

3:0.10

4:0.07

5:0.04

6:0.15

7:0.31

Negative Logits

WriteBarrier

-0.45

parsedMessage

-0.44

DebuggerNonUser

-0.43

WithIOException

-0.41

原始内容存档于

-0.41

LookAnd

-0.41

 protoimpl

-0.40

 aapt

-0.40

Попис

-0.40

 ProtoMessage

-0.39

POSITIVE LOGITS

 Grüße

0.24

formKey

0.22

skosten

0.22

dó

0.22

épa

0.22

例子

0.21

volkswagen

0.21

 Automat

0.21

haf

0.21

 Rück

0.21

Activations Density 0.008%

attends to the token "label" from tokens of various markers

No Comments

No Known Activations