INDEX

Explanations

phrases related to rejection and condemnation

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

block restaurant openings

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

 Kost

-0.43

 quarter

-0.38

 loud

-0.38

 elegant

-0.37

 restauration

-0.35

 nodding

-0.34

 Elegant

-0.34

 stress

-0.34

Why

-0.33

 depth

-0.33

POSITIVE LOGITS

 defaultstate

0.59

UnknownFields

0.55

şört

0.52

RTEE

0.52

PreferredItem

0.51

yelitis

0.51

 unknownFields

0.50

trab

0.50

 boycot

0.48

 AssemblyTitle

0.48

Activations Density 0.035%

phrases related to rejection and condemnation

block restaurant openings

No Comments

No Known Activations