INDEX

Explanations

expressions of belief, suspicion, and perception regarding personal or others' experiences

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

think, believe, suspect

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

 ciò

-0.32

 Completion

-0.31

 actionPerformed

-0.31

Completion

-0.31

ad

-0.30

Propel

-0.30

了口气

-0.29

ことに

-0.29

freopen

-0.29

 here

-0.29

POSITIVE LOGITS

ftagPool

0.65

KommentareTeilen

0.63

 AssemblyTitle

0.60

Cyfarwyddwr

0.59

OGND

0.55

 intStringLen

0.54

ConstraintMaker

0.53



0.52

 Taktlose

0.52

 jsPsych

0.51

Activations Density 0.038%

expressions of belief, suspicion, and perception regarding personal or others' experiences

think, believe, suspect

No Comments

No Known Activations