INDEX

Explanations

numerical values, particularly those related to measurements, references, and items in a list

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

figure and equation references

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

 externi

-0.56

تفصیلات

-0.48

curities

-0.47

brides

-0.47

pecified

-0.47

ější

-0.46

 savior

-0.46

 ſind

-0.46

 Corbin

-0.46

 linkovi

-0.45

POSITIVE LOGITS

 CreateTagHelper

0.52

:][

0.46

 '\\;'

0.44

 réus

0.44

Personendaten

0.43

 surla

0.40

 AppCompatTheme

0.39

 iniziale

0.38

ViewFeatures

0.36

 îna

0.36

Activations Density 0.010%

numerical values, particularly those related to measurements, references, and items in a list

figure and equation references

No Comments

No Known Activations