INDEX

Explanations

attends to various tokens that are indicative connective phrases from preceding 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

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.07

1:0.10

2:0.09

3:0.12

4:0.27

5:0.17

6:0.08

7:0.07

Negative Logits

-4.97

-4.28

↵↵

-4.22

-4.16

-3.92

<eos>

-3.92

-3.88

↵

-3.53

in

-3.50

POSITIVE LOGITS

 myſelf

8.88

 itſelf

8.56

Efq

8.44

 Monfieur

8.13

 Theſe

7.69

 Jefus

7.63

 Majefty

7.50

 ―――――

7.50

 pleaſure

7.50

 purpoſe

7.44

Activations Density 0.026%

attends to various tokens that are indicative connective phrases from preceding tokens

No Comments

No Known Activations