INDEX

Explanations

quantifiers and expressions related to extent or degree

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

so far

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

 cardiaque

-0.38

 devint

-0.35

 térmico

-0.34

 grecque

-0.33

 gué

-0.33

 brukes

-0.33

 тъ

-0.32

 noires

-0.32

 plegable

-0.32

 véritable

-0.31

POSITIVE LOGITS

fjspx

0.80

rungsseite

0.77

 bisher

0.73

 sejauh

0.72

为止

0.72

Efq

0.71

featureID

0.70

 المعيارى

0.70

setVerticalGroup

0.69

dafx

0.67

Activations Density 0.015%

quantifiers and expressions related to extent or degree

so far

No Comments

No Known Activations