INDEX

Explanations

attends to language-related tokens marked as gloss from pairs of language-specific tokens

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 0-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_0/width_16k/average_l0_104

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.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

0:0.18

1:0.18

2:0.12

3:0.19

4:0.11

5:0.06

6:0.05

7:0.07

Negative Logits

 myſelf

-2.67

 itſelf

-2.61

 Monfieur

-2.56

Efq

-2.52

 purpoſe

-2.36

 pleaſure

-2.34

 Majefty

-2.31

 Anſ

-2.31

 Theſe

-2.31

 Jefus

-2.30

POSITIVE LOGITS

1.44

1.24

in

1.17

↵

1.08

to

1.06

the

1.04

1.02

1.01

Activations Density 0.000%

attends to language-related tokens marked as gloss from pairs of language-specific tokens

No Comments

No Known Activations