INDEX

Explanations

attends to tokens that are followed by a specific target token marked as "leuco."

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

0:0.72

1:0.01

2:0.02

3:0.02

4:0.11

5:0.03

6:0.01

7:0.04

Negative Logits

<eos>

-0.53

-0.51

-0.47

in

-0.47

-0.46

-0.45

to

-0.44

of

-0.43

POSITIVE LOGITS

 itſelf

0.88

 myſelf

0.87

Efq

0.81

 Jefus

0.79

 ſtate

0.77

ſelf

0.75

 Theſe

0.75

 Majefty

0.74

 Houſe

0.74

 raiſ

0.74

Activations Density 1.629%

attends to tokens that are followed by a specific target token marked as "leuco."

No Comments

No Known Activations