INDEX

Explanations

expressions indicating causality or conditions

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

connecting phrases and concepts

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

 pleaſure

-0.58

 propOrder

-0.57

 NSCoder

-0.57

MemoryWarning

-0.55

 disambiguazione

-0.54

OGND

-0.52

 Goliath

-0.52

ynb

-0.51

 vegans

-0.51

 feminist

-0.51

POSITIVE LOGITS

Autoritní

0.46

 nisso

0.36

:✨

0.36

 PUSTAKA

0.35

OCCURRED

0.34

 <<<<<<<<<<<<<<

0.34

hafte

0.33

toarele

0.32

 antworte

0.32

ţion

0.32

Activations Density 0.106%

expressions indicating causality or conditions

connecting phrases and concepts

No Comments

No Known Activations