INDEX

Explanations

syntax elements commonly found in programming code

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

technical and older language

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

 reversed

-0.36

exitRule

-0.33

 favor

-0.31

 reversing

-0.31

 calldata

-0.30

formik

-0.30

 Allie

-0.30

 scurt

-0.30

eqn

-0.30

 estimation

-0.29

POSITIVE LOGITS

SpringBootTest

0.68

 ddelweddau

0.68

 pleaſure

0.64

 houſe

0.63

ſelves

0.60

 myſelf

0.60

 pinulongan

0.59

 EconPapers

0.58

 purpoſe

0.58

 Jefus

0.56

Activations Density 0.000%

syntax elements commonly found in programming code

technical and older language

No Comments

No Known Activations