INDEX

Explanations

quantitative survey results and statistical findings

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

examples and findings

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

featureID

-0.55

!*\

-0.52

"":

-0.52

ꞌ

-0.50

HandlerContext

-0.50

⃗

-0.50

`"

-0.50

hone

-0.49

UCI

-0.49

VersionUID

-0.49

POSITIVE LOGITS

 infatti

0.45

 addirittura

0.41

 misalnya

0.41

 namorado

0.41

 sogar

0.40

rungsseite

0.39

 jopa

0.39

 nawet

0.39

 notamment

0.38

 zelfs

0.38

Activations Density 0.148%

quantitative survey results and statistical findings

examples and findings

No Comments

No Known Activations