INDEX

Explanations

attends to the pattern of tokens indicating equality from tokens suggesting inequality

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

0:0.05

1:0.07

2:0.06

3:0.06

4:0.06

5:0.03

6:0.35

7:0.28

Negative Logits

 AspNetCore

-0.38

ponses

-0.26

这位

-0.25

Etc

-0.25

évaluateur

-0.24

<<(

-0.24

udsman

-0.24

 ricordare

-0.23

ernos

-0.23

Etc

-0.23

POSITIVE LOGITS

Abitanti

0.42

AntiForgeryToken

0.39

amssymb

0.36

resizingMask

0.35

 بيها

0.35

Билгалдахарш

0.35

ANTLR

0.34

 nakalista

0.34

 وتسجيلات

0.33

 Linton

0.33

Activations Density 0.387%

attends to the pattern of tokens indicating equality from tokens suggesting inequality

No Comments

No Known Activations