INDEX

Explanations

specific phrases or references that denote credibility or reliability in information

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

numbers and measurements

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

tokens that appear in technical or formatted contexts (code identifiers, XML/tags, section-heading or list-intro words and other emphasized/structural document tokens).

oai_token-act-pair · gpt-5-mini Triggered by @vetterc0

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



-0.50

:✨

-0.45

 otomatig

-0.45

RegressionTest

-0.44

URLException

-0.44

 препратки

-0.44

 Infórmanos

-0.43

ViewImports

-0.43

Nuorodos

-0.43

Inflater

-0.43

POSITIVE LOGITS

principalTable

0.46

 Khusus

0.43

 vocación

0.37

disposing

0.37

Trans

0.36

abstractmethod

0.34

ordnung

0.34

 cambiado

0.33

Torn

0.33

wią

0.33

Activations Density 0.296%

specific phrases or references that denote credibility or reliability in information

numbers and measurements

tokens that appear in technical or formatted contexts (code identifiers, XML/tags, section-heading or list-intro words and other emphasized/structural document tokens).

No Comments

No Known Activations