INDEX

Explanations

attends to tokens related to "launch" from tokens indicating a subsequent action or change, often related to operations or events being 'shut down.'

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-att/layer_0/width_16k/average_l0_61

Prompts (Dashboard)

16,384 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

0:0.01

1:0.01

2:0.55

3:0.03

4:0.00

5:0.00

6:0.06

7:0.03

8:0.01

9:0.01

10:0.01

11:0.01

12:0.01

13:0.11

14:0.07

15:0.01

Negative Logits

<unused43>

-0.55

<unused41>

-0.54

<unused42>

-0.54

<unused47>

-0.54

<unused23>

-0.54

[@BOS@]

-0.54

<unused16>

-0.54

<unused17>

-0.54

<unused3>

-0.54

<pad>

-0.54

POSITIVE LOGITS

is

0.15

of

0.15

se

0.12

lo

0.12

can

0.12

re

0.12

 would

0.11

may

0.11

Activations Density 0.055%

attends to tokens related to "launch" from tokens indicating a subsequent action or change, often related to operations or events being 'shut down.'

No Comments

No Known Activations