INDEX

Explanations

specific names and data elements, particularly related to personal identifiers and numerical values

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

numbers and mathematical operations

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

st

-0.39

 esterni

-0.37

si

-0.36

 come

-0.34

 Pende

-0.33

-0.32

 extra

-0.32

 leading

-0.32

 setuptools

-0.32

enschappelijke

-0.32

POSITIVE LOGITS

enumii

0.81

 betweenstory

0.80

ſelf

0.77

 myſelf

0.73

 raiſ

0.72

 Monfieur

0.71

 purpoſe

0.70

 becauſe

0.70

 deſt

0.69

 ainfi

0.68

Activations Density 0.360%

specific names and data elements, particularly related to personal identifiers and numerical values

numbers and mathematical operations

No Comments

No Known Activations