INDEX

Explanations

questions and phrases related to actions and tasks

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

asking or describing actions

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

 vertrou

-0.47

 cadrul

-0.45

 desconhe

-0.45

 páss

-0.45

 dezelve

-0.44

 pescoço

-0.43

 mijne

-0.43

 magnétique

-0.42

bibfield

-0.42

 magnética

-0.41

POSITIVE LOGITS

 doings

0.73

 DONE

0.60

doing

0.59

 DOING

0.57

happen

0.57

 Aktion

0.56

 Done

0.56

happens

0.56

CloseOperation

0.56

 done

0.55

Activations Density 0.014%

questions and phrases related to actions and tasks

asking or describing actions

No Comments

No Known Activations