INDEX

Explanations

words related to actions of evaluation or legal terms

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

instructed, intervened, supplemented, warranted, warned, licensed

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

 незавершена

-0.77

fjspx

-0.62

extAlignment

-0.60

 pinulongan

-0.54

YourGuide

-0.53

StoreMessageInfo

-0.51

Diwedd

-0.51

ddots

-0.51

GTCX

-0.50

OrWhiteSpace

-0.50

POSITIVE LOGITS

Personendaten

0.40

Aiheesta

0.35

strijden

0.35

 ***!

0.34

prepareStatement

0.33

 differently

0.33

 defend

0.33



0.32

?”

0.32

RuleContext

0.32

Activations Density 0.063%

words related to actions of evaluation or legal terms

instructed, intervened, supplemented, warranted, warned, licensed

No Comments

No Known Activations