INDEX

Explanations

attends to negation tokens from affirmative assertions or questions

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

0:0.07

1:0.08

2:0.13

3:0.06

4:0.03

5:0.01

6:0.07

7:0.51

Negative Logits

 distanciation

-0.31

&__

-0.28

PositiveButton

-0.28

tioner

-0.26

 Toen

-0.26

IBOutlet

-0.26

}_{

-0.25

icago

-0.24

 enrollment

-0.24

拮

-0.24

POSITIVE LOGITS

WriteBarrier

0.36

/**

0.35

setVerticalGroup

0.35

 SwitchCompat

0.34

WireFormatLite

0.33

 gärna

0.32

 CURIAM

0.32

FormTagHelper

0.31

 Lumpur

0.31

↑

0.31

Activations Density 1.464%

attends to negation tokens from affirmative assertions or questions

No Comments

No Known Activations