INDEX

Explanations

attends to the pronouns and related expressions from the previous tokens that refer to people or elements in the discussion

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

0:0.06

1:0.08

2:0.06

3:0.10

4:0.08

5:0.03

6:0.44

7:0.11

Negative Logits

WebVitals

-0.34

Espèce

-0.30

WriteBarrier

-0.30

 مشين

-0.29

Interviewer

-0.28

|')

-0.28

uxxxx

-0.28

 mostly

-0.27

CompleteListener

-0.27

)|^{

-0.27

POSITIVE LOGITS

 كومونز

0.41

hazi

0.37

 ویکی‌پدی

0.35

 autorytatywna

0.35

Micah

0.34

 JADX

0.33

tvguidetime

0.33

miyorum

0.33

 Micah

0.33

ξης

0.32

Activations Density 0.004%

attends to the pronouns and related expressions from the previous tokens that refer to people or elements in the discussion

No Comments

No Known Activations