INDEX

Explanations

code-related terminology and syntax elements

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

test file paths and names

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

AndEndTag

-0.57

RotationOrder

-0.47

請繼續往下閱讀

-0.43

xically

-0.38

hassee

-0.33

السكان

-0.33

ContentAsync

-0.33

 становника

-0.33

ніципа

-0.32

<bos>

-0.32

POSITIVE LOGITS

 bescha

0.47

quillo

0.46

domada

0.46

 Administrativna

0.45

ppuden

0.45

utilisons

0.44

 Sumatera

0.44

 Irvin

0.44

httphttps

0.44

Demo

0.44

Activations Density 0.034%

code-related terminology and syntax elements

test file paths and names

No Comments

No Known Activations