INDEX

Explanations

phrases and terms related to types of letters (uppercase and lowercase) and names of specific characters or entities

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

Followed by a letter

np_acts-logits-general · gemini-2.0-flash

single letters and vitamins

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

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

•Gemma-2-2B-L20-Structure-Cluster

Technical Natural Language (modifiers of or English-based letter categories)

No Comments

Negative Logits

ſelf

-0.94

EndGlobalSection

-0.93

 חיצוניים

-0.87

 calendriers

-0.85

ChildScrollView

-0.79

neſs

-0.78

CppMethod

-0.76

 ProtoMessage

-0.76

DockStyle

-0.76

 Diſ

-0.76

POSITIVE LOGITS

1.28

1.23

1.19

1.14

1.12

1.09

1.08

1.02

Activations Density 0.992%

phrases and terms related to types of letters (uppercase and lowercase) and names of specific characters or entities

Followed by a letter

single letters and vitamins

No Comments

No Known Activations