INDEX

Explanations

sections of a document that structure information

oai_token-act-pair · gpt-4o-mini Triggered by @bot

section headers

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 мәкал

-0.54

 ब्रेकडाउन

-0.54

</b>

-0.46

رشف

-0.43

 للمعارف

-0.39

 вовсе

-0.38

Erreferentziak

-0.38

министра

-0.38

IVEREF

-0.37

 disambiguazione

-0.36

POSITIVE LOGITS

setVerticalGroup

0.64

+#+#

0.54

StoryboardSegue

0.52

 propOrder

0.52

uxxxx

0.47

XmlAccessType

0.47

.*")]

0.47

 ſta

0.42

 paſſ

0.42

 ſtand

0.42

Activations Density 0.122%

sections of a document that structure information

section headers

No Comments

No Known Activations