INDEX

Explanations

expressions of positive sentiment and approval

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

words of strong approval

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

__":

-0.57

esgue

-0.51

 noDo

-0.50

 FetchType

-0.48

RectangleBorder

-0.48

edance

-0.47

Rhestr

-0.46

semble

-0.45

kaŭ

-0.44

byshev

-0.43

POSITIVE LOGITS

yesterday

0.44

 occaf

0.43

 pleaſure

0.43

 duquel

0.42

UnusedPrivate

0.42

 Yesterday

0.42

 anſ

0.40

createStatement

0.39

 raiſ

0.39

 régal

0.39

Activations Density 0.013%

expressions of positive sentiment and approval

words of strong approval

No Comments

No Known Activations