INDEX

Explanations

technical terms related to adjustments and corrections in research data

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

corrected, adjusted, shifted

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

 للمعارف

-0.69

expandindo

-0.68

AddTagHelper

-0.67

 indígen

-0.66

 Geſ

-0.64

ſehen

-0.64

 témoig

-0.63

orianCalendar

-0.63

 Anſ

-0.63

 ddelwed

-0.59

POSITIVE LOGITS

 "..\..\

0.32

tun

0.31

 Насе

0.29

Locus

0.28

我看

0.28

 womit

0.27

 makes

0.27

rid

0.27

 peti

0.27

makes

0.26

Activations Density 0.266%

technical terms related to adjustments and corrections in research data

corrected, adjusted, shifted

No Comments

No Known Activations