INDEX

Explanations

terms related to programming libraries and modules

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

filenames and paths

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

 hänen

-0.25

 demikian

-0.24

GeneratedMessage

-0.24

 wikipagina

-0.23

ถม

-0.23

OrBuilder

-0.23

 parezca

-0.22

 pungkas

-0.22

 nakalista

-0.22

 furent

-0.20

POSITIVE LOGITS

setVerticalGroup

0.81

ዡ

0.77

𑄥

0.77

娡

0.77

𑄮

0.76

<unused41>

0.76

<unused14>

0.75

<unused1>

0.75

[@BOS@]

0.75

<unused17>

0.75

Activations Density 0.010%

terms related to programming libraries and modules

filenames and paths

No Comments

No Known Activations