INDEX

Explanations

expressions emphasizing totality or completeness

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

all and french/spanish equivalents

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

 febru

-0.48

VPS

-0.47

Rptr

-0.47

erdings

-0.45

PDC

-0.45

SDC

-0.45

SQLException

-0.45

MPN

-0.45

VPS

-0.44

 Jacobsen

-0.43

POSITIVE LOGITS

all

0.84

All

0.77

All

0.75

ALL

0.73

all

0.73

Toda

0.72

 Toda

0.71

 everything

0.71

 segala

0.71

Toute

0.71

Activations Density 0.002%

expressions emphasizing totality or completeness

all and french/spanish equivalents

No Comments

No Known Activations