INDEX

Explanations

sequences of whitespace characters

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

`And article`

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

Negative Logits

iſchen

-0.91

RegressionTest

-0.89

<unused41>

-0.89

ſelben

-0.88

<unused74>

-0.88

<unused68>

-0.88

<unused23>

-0.88

<unused8>

-0.88

<unused3>

-0.88

<pad>

-0.88

POSITIVE LOGITS

0.33

0.31

bal

0.30

0.29

En

0.28

Can

0.28

Activations Density 0.002%

sequences of whitespace characters

`And article`

No Known Activations