INDEX

Explanations

recurring patterns or elements related to coding and markup language syntax

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

web code snippets

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

 GenerationType

-0.66

 itſelf

-0.62

 pleaſure

-0.57

 ſta

-0.56

IntoConstraints

-0.56

featureID

-0.56

 nakalista

-0.54

 ſtand

-0.54

 indígen

-0.54

 ſte

-0.53

POSITIVE LOGITS

 shit

0.41

猫

0.40

 modeling

0.39

 mumbled

0.39

lol

0.39

mom

0.39

 stuff

0.37

 cats

0.36

vô

0.36

 shitty

0.36

Activations Density 0.034%

recurring patterns or elements related to coding and markup language syntax

web code snippets

No Comments

No Known Activations