INDEX

Explanations

conjunctions or connecting phrases in the text

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

phrases of connected descriptors

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

foglal

-0.45

 fırs

-0.42

 vigencia

-0.40

 smlou

-0.40

 ditangkap

-0.39

either

-0.39

 montanha

-0.39

 vrá

-0.39

nji

-0.38

 poursu

-0.38

POSITIVE LOGITS

 highly

0.77

 unſ

0.76

 extremely

0.74

ftagPool

0.71

 فريبيس

0.70

 EconPapers

0.65

 Eilish

0.65

 myſelf

0.64

IntoConstraints

0.64

contentLoaded

0.63

Activations Density 0.039%

conjunctions or connecting phrases in the text

phrases of connected descriptors

No Comments

No Known Activations