Latest Jaunts

@yeeguy

3 Followers

TerraFuture Seed Syndicate r8.pptx

TerraFuture

A Portfolio Approach to Invest in Nature-Based Solutions Opportunities , starting in Africa

Our Planet Our Future





1

0



0



34

#carboncredits #sustainablefinance #naturebasedsolutions #clima

22 hours ago

@yeeguy

3 Followers

Terraformation Workshop at NOAH 2023.pptx

Terraformation

NOAH 2023

Carbon Forestry as a

Securitizable Asset Class

Yee Lee





1



0

yee@terraformation com

0

#forests #nbs #carbonremovals #terraformation

22 hours ago

@yeeguy

3 Followers

Terraformation at NOAH24.pptx

Terraformation

NOAH 2024

Yee Lee

yee@terraformation com

Terraformation





0

0



0

#forests #nbs #carbonremovals #terraformation

22 hours ago

@roshni

3 Followers

World War-2 Quiz

World War II Quiz

Roshni Ranjan



0

@yeeguy

+

FOLLOW

terrafuture org

verponts

Terraformation



2



@roshni

This is a quiz on World War-2 I have made.

Lets see how much you know about the World War-2! ... m

#history #historyquiz #trivia #adolfhitler #worldwar

1 day ago

@roshni

3 Followers

My Sketching and Drawing Album

MY SKETCHING & PAINTING ALBUM

BY ROSHNI RANJAN





2

@roshni

These are some of the sketches and paintings I

have created.

Take a stroll through my ... more

See more comments

#landscape #sketching #drawing #creativeexpression #painting

1 day ago

@roshni

3 Followers

Planets Quiz





2

@roshni

This is a quiz I created for the students in my

hobby class. Let's see how much you know!

#spacequiz #solarsystem #educationalchallenge #astronomyfun #

1 day ago

@roshni

3 Followers

Dandelions - A Poem

Dandelions

One Sunday morning,

After a light shower of rain

A small, beautiful dandelion rose in the air

It was different

the others, pure white

from

It glistened in the sunlight, sunny and bright!

The small dandelion was dancing around

All the other dandelions prancing around

Suddenly; a fierce wind picked up the air

The dandelion was angry at

separated from its friends.

being

It started to thunder and then came a rumbling sound

Thunder crashed, and it started to rain

The dandelion was swept away!

The wind whirled it around and rain poured water on it.

again!

The poor dandelion was sad

Then the dandelion saw a ray of light!





3

@roshni

Thank you!

See more comments

#childrensliterature #dandelions #adventure #joy #naturepoetry

1 day ago

@roshni

3 Followers

Conservation Of Forests

CONSERVATION OF

FORESTS

and

Devanshi Jain

+

FOLLOW



27

3



0

+

FOLLOW



37

1



+

FOLLOW



27

2



0



2



@roshni

This is a project I made with my classmates on

Conservation of Forests.

The students who ... more

#a

"

orestation #enviroment #climatechange #wildlifeconservation

1 day ago

@roshni

3 Followers

Spreading Awareness on Mental Health

MENTAL

HEALTH

BY: ROSHNI

RANJAN

TRISHA BANERJEE

KRITI IYER





2

@roshni

This is a presentation that I have made with my

classmates on Mental Health. The students who have ... m

#mentalhealthawareness #health #depression #mentalhealth #me

1 day ago

@jonathan

21 Followers

Leveraged ETFs vs Margin Loans: why borrowing is

better

Leveraged ETFs vs. Margin Loans: Why Borrowing is Often Better

Leveraged ETFs, like TQQQ (3x leveraged Nasdaq-100) and UPRO (3x leveraged S&P

glance, they seem like a safer, simpler alternative to borrowing money to invest.

500), offer an easy way to amplify returns without taking on a margin Ioan. At first

However, when you compare the two approaches, margin loans often outperform

leveraged ETFs over time

The reason lies in how leveraged ETFs are structured, and

erode returns

how their mechanics_like beta slippage, roll yield, and tracking errors

To illustrate, lets compare the outcomes of investing $100 using a leveraged ETF

versus borrowing $100 to double your exposure to the underlying index.

The Scenario: A Volatile Market

Imagine an index starts at $100, rises 10% to $110, and then falls back 10% to $99.

We'll look at the outcomes for a 3x leveraged ETF and a margin loan over these two

Leveraged ETF (3x Daily Performance)

Day

goes from $100 to $130_

Day

1: The index rises from $100 to $110 (+10%). The ETF gains 30%, so it

2: The index falls from $110 to $99 (-10%). The ETF loses 30%, so it drops

from $130 to $91.

After two days: The index is down

% , but the leveraged ETF is down 9%.

Using a Margin Loan (2x Leverage)

With a margin loan, you invest $200 total-~your $100 plus $100 borrowed. If the index

your $220 becomes $198_

After two days: Your portfolio is down 1%, matching the index.

Why Borrowing Outperforms Leveraged ETFs

1. Beta Slippage: The Compounding Problem

Leveraged ETFs are designed to multiply the daily returns of the index, not its

long-term performance _

When the index rises and falls, the compounding effect works

9% while the index fell only 1%, a dramatic divergence caused by this daily

against the ETF over time, particularly in volatile markets. In our example, the ETF fell

compounding effect. With a margin loan, your returns directly track the index's

performance, so you avoid this slippage.





0

0



0



43

#financeeducation #marginloans #investmentstrategy #financiala

1 day ago

@jonathan

21 Followers

Investing strategy for individual equities

Investing for the Long Haul: A Strategy to Minimize Decisions

and Maximize Returns

Investing is 90% a game of psychology: The biggest challenge isn't

One way to master this is by creating a strategy that minimizes the

number of decisions you need to make. Fewer decisions mean fewer

opportunities for mistakes, and in the long run, this can translate to

companies, building small initial positions, and protecting my

better outcomes. My approach focuses on selecting dominant

psychology while letting compounding do the heavy lifting

Ready to post

Complete file details

Title (Required)

Jaunt



Category

Business

#Tags (Use # for tags, separate with space)

Summary (upto 500 characters)





0

0



0



25

#emotionalintelligence #longterminvesting #investingstrategy #in

1 day ago



Let AI suggest a summary

This presentation explores innovative

platforms that optimize dietary

requirement gathering for food services.

@jonathan

21 Followers

An American Life

It highlights how these platforms reduce

Add a comment

M

What's on your mind with this

doc/deck? (shows up as first

comment)



PUBLISH YOUR POST

AN

A MERIGAN

FAMILY





0



0

0



38

#culturalnarrative #americanheritage #personalstory #familyhistor

1 day ago

@jonathan

21 Followers

An American Life

AN

A MERIGAN

FAMILY





0



0

0



26

#heritagestories #mythsandtruths #boutellelegacy #familyhistory

1 day ago

@Vikram

5 Followers

Digestive system

The Digestive

System!

By; Josh, Sameera, and Vikram

1





0



0



81

#biology #humanbody #digestivesystem #nutrition #health

4 days ago

@jonathan

21 Followers

Material tranforms from disentangled NeRF

representations

Material transforms from disentangled NeRF representations

Ivan Lopes

Jean-François Lalonde?

'nria

Raoul de Charette

2Université Laval

novelviews

Transfer BRDF transform to new scenes

complex

o

scenes. From

observations of a scene in nwo conditions: original and transformed, we leverage a joint Neural Radiance Field (NeRF ) optimization to learn

Leamned

1

Learnlng BRDF transform from multivlews

Figure I: Proposed method We illustrate our approach for inferring unknown material transformations in

material mapping function F which models the observed changes at the material level accurately (e.g the topmost transform on the left is

Abstract

In this paper; we first propose

novel method for

disentangled Neural Radiance Field (NeRF) representations, our approach learns to map Bidirectional Reflectance Distribution

transferring material transformations across different scenes. Building on

Functions (BRDF) from pairs of scenes observed in varying conditions, such as dry and wet. The learned transformations can

then be applied to unseen scenes with similar materials,therefore effectively rendering the transformation learned with an

arbitrary level of intensity Extensive experiments on synthetic scenes and real-world objects validate the effectiveness of our

only the versatility of our method but also its potential for practical applications in computer graphics. We publish our method

implementation, along with our synthetichreal datasets on https:/ /github.com/astra

Introduction

key

In computer graphics and vision, inverse rendering is

ing

toextract-

material information and allowing re-rendering under novel

While neural repre-

sentations have largely taken over the traditional Physically-Based

conditions (viewpoint, lighting, materials, etc.)

Rendering (PBR) techniques, recent works have demonstrated that

presery-

the two representations can be combined [JLX* 23]. thus

ing

the editability and expressivity of PBR representations along

with the flexibility of neural representations.



1



0



1 week ago

@jonathan

21 Followers

Incentive Design With Spillovers

INCENTIVE DESIGN

KRISHNA DASARATHA

ABSTRACT.

principal

uses

WITH SPILLOVERS

BENJAMIN GOLUB. AND ANANT SHAH

payments conditioned

stochastic outcomes of a team

project to elicit costly effort from the team members. We

ization of a classic first-order

from network games.

tive pay

across

agents; a

'develop

multi-agent general-

approach to contract optimization by leveraging methods

The main results characterize the optimal allocation of incen-

agents and outcomes.

Incentive optimality requires equalizing; across

product of (i) individual productivity (ii) organizational centrality and (iii)

responsiveness to monetary incentives.

Dasaratha:

western

krishnadasaratha@

University:

to Drew

Shah:

Fudenberg and his reading group for a very valuable discussion_

ben golub @gmail com,

North-

We thank

We are

and also for helpful comments

Omer Tamuz, Marina Halac, Thomas Steinke, Marzena Rostek, Daron Ravid, Jose Betancourt

to Ilya Segal,

Mailath;

George

lencia, Roberto Corrao, Jacopo Perego; Juan Ortner,

Gottardi, and Michael Powell,



E

1



0

0



grateful

Va-

Adam Szeidl, Sam Wycherley; Michael Ostrovsky; Elliot Lipnowski, Yifan Dai, Melika Liporace; Piero

as well

0



84

1 week ago

@kapilmohan

9 Followers

Can AI Sniff Like Humans?

SNIFF AI: Is MY

YOUR

SPICY

PERCEPTUAL ALIGNMENT WITH HUMAN SMELL ExPERIENCES

Shu

Zetao Zhou? , Christopher Dawes"

Giada Brianz

and Marianna Obrist

'Zhong

Department of Computer Science, University College London; United Kingdom

2Division of Psychology and Language Sciences, University College London; United Kingdom

Further erploration analyses on how

Al claude

lavender: (The scent is floral

and fresh with a subtle sweetness and

herbal undertones.Ithas

calming and

soothing quality that can evoke feelings

of relaxation and tranquility.The aroma is

clean and

light, reminiscent of

breeze passing through

on

garden ,

The relerence kent

gentle

Alex Wolitzky;

Jason Hartline,

as many seminar and conference participants.

'gmail

0



143

When considering the appearance of

scene

coat of varnish)

certain transfor-

canalter the ma-

scene

change drastically . Currently, estimating the PBR characteristics of

known material after such

scene again in the desired target condition. This process is both

and

complex

transformation requires capturing the

laborious due t0 the variety of possible transforma -

tions; such as wetness, dust, varnish, painting. etc. In this work

aim to lcarn

BRDF transformation from

it to different scenes.

source scene and

WC

apply

MYTHS

HALF

Will Boutelle

LIFE

TRUTHS

MYTHS

TRUTHS

LIFE

& HALF TRUTHS

Will Boutelle

1



0

3

g

1

5

g

1



0

quite

this Smell. The Sent

more

fowery and Jweet compared

Smells

does

which

Smells _

I: SniffAl: an overview of user study and examples from the user study.

Figure

ABSTRACT

Aligning Al with human intent is important, yet perceptual alignment

how Al interprets what

on

This work focuses

user study with 40 participants to investigate how well AI can interpret

olfaction; human smell

interactive tasks. with our

hear, or smell-

experiences.

human

designed Al system attempting to guess what scent the participants were experiencing based on their

remains underexplored.

We conducted

descriptions of scents. Participants performed "sniff and describe'

descriptions. These tasks evaluated the Large Language Model's (LLMs) contextual understanding

and representation of scent relationships within its internal states

space:

Results indicated limited perceptual alignment, with biases towards certain scents, like lemon and

Both quantitative and qualitative methods were used to evaluate the Al system

failing

peppermint, and continued

high-dimensional embedding

identify others, like rosemary. We discuss these

HCI systems with multisensory experience integration.

performance.

'findings in light

alignment advancements; highlighting the limitations and opportunities for enhancing

Introduction

Aligning Artificial intelligence (Al) behaviour with human preference is critical for the future of AL An important yet

between Alassessments and human

often overlooked aspect of this alignment is the perceptual alignment . Perceptual alignment refers to the agreement



0



2 weeks ago

@jonathan

21 Followers

SMoA: Improving Multi-agent Large Language

Models with Sparse Mixture-of-Agents

SMoA: Improving Multi-agent Large Language Models with

Sparse Mixture-of-Agents

Dawei Li'

Kumar Satvik Chaudhary

Zhen Tan' , Peijia Qian? , Yifan Li' ,

'School of Computing, and Augmented Intelligence, Arizona State University

Lijie Hu' , Jiayi Shens

2Independent Researcher

Abdullah University of Science and Technology

'Computer Science and Engineering, Michigan State University

SUniversity of Texas at Austin

'King

Abstract

multi

While

significantly enhance the performance of

agent systems have been shown

Large Language Models (LLMs)

across

var-

ious tasks and applications, the dense interac -

tion between

pers their efficiency and diversity. To address

scaling

these challenges,

agents potentially ham-

inspiration from

pose a sparse mixture-of-agents (SMoA) frame -

the sparse mixture-of-agents (SMoE) and pro -

work to improve the efficiency and diversity

of multi-agent LLMs. Unlike completely con-

nected structures, SMoA introduces novel Re-

sponse Selection and Early Stopping mecha-

nisms to sparsify information flows among in-

dividual LLM agents; striking a balance be-

tween

ally; inspired by the expert diversity principle

in SMoE frameworks for workload balance

between experts.

we

scriptions to each LLM agent, fostering diverse

and

On

reasoning, alignment, and fairness bench-

marks demonstrate that SMoA

formance

agents approaches but with significantly lower

comparable to traditional mixture-of-

computational costs. Further analysis reveals

that SMoA

pacity

is more stable, has a greater

to scale,

ca-

and offers considerable po-

tential through hyper-parameter optimization.

Code and data will be available

at:

Ilgithub. com/David-Li0406/SMoA.

Introduction

The rapid development of Large Language Models

(LLMs) (Brown et al., 2020; Anil et al., 2023; Gan

et al., 2023; Dubey et al., 2024) in recent years

has significantly advanced in a series of NLP tasks;

such

Tong

as

et al., 2023; Jin et al., 2024a), knowledge dis-

covery (Li et al , 2024a,b; Tan et al , 2024a) and

While these powerful foundation models have been

question answering (Wang et al., 2022;

systems (Li et al, 2022; Sun et al., 2024).

dialogue

proven to benefit from extensive training data and



1

1



0

0



2 weeks ago

@jonathan

21 Followers

Causal Responsibility Attribution for Human-AI

Collaboration

Causal Responsibility Attribution for Human-AI Collaboration

Yahang Qi

ETH Zurich

Bernhard Schölkopf

MPI

Tübingen

Zhijing Jin

University of Toronto

YAHAQI@ETHZ.CH

BS@TUEBINGEN .MPG.DE

ZJIN@CS.TORONTO.EDU

0



92

https:

assign distinct role de-

divergent thinking. Extensive experiments

achieves per-

Prompt

Output

Multi-agent debate

Output

MoA

Output

SMoA (ours)

Figure I: Comparison among the pipeline structures of

MAD, MoA and SMoA

larger model sizes, further scaling up these mod-

retraining

els is exceptionally costly, often requiring necessi -

several trillion tokens (Zhang

tating

et al., 2024). To overcome this limitation, multi -

agent LLMs (Liang et al , 2023;

et al , 2024b)

have been explored, enabling LLM systems to in-

Wang

corporate multiple agents; each focused on distinct

objectives and tasks.

The

Wang

based structure

(Liang et al., 2023;

et al., 2024b) is one of the

layer-|

multi-agent systems.

various

LLMs

to

most funda-

involves

perform

(MAD) (Liang et al., 2023; Du et al., 2023) or dis-

multi-agent debate

cussion over multiple rounds to mimic human be-

havior in problem-solving.

While effective in vari -

ous applications, early layer-based methods (Li

agent each time, hindering their utilization in

et al., 2023) process user queries with only one

real-

(MoA)

-world scenarios. Recently, mixture-of-agents

et al., 2023;

et al., 2024b) has

been devised, which uses multiple processors per

(Zhang

Wang

layer to handle queries simultaneously for time ef-

these references to produce a final answer:

ficiency purposes. An aggregator then synthesizes

Although MoA improves the time efficiency of

multi-agent LLMs, it still faces

significant chal -

lenges. The first issue is the high token computa-

tional cost. While MoA reduces processing time

instructing

subjective judgments across different sensory modalities, such as vision, hearing,

0



65

ž

1

1

1



0

in critical sectors

such as healthcare (Budd et al , 2021), finance (Cohen et al, 2023), and autonomous driving (Badue

Abstract

As Artificial Intelligence (Al) systems increasingly influence decision-making across various fields,

the need to attribute responsibility for undesirable outcomes has become essential, though compli -

cated by the complex interplay between humans and AL

actual

to an outcome and rely on real-world measures of blameworthiness that may misalign with respon -

Existing attribution methods based

on

causality and Shapley values tend to disproportionately blame agents who contribute more

sible Alstandards. This paper presents a causal framework

Structural Causal Models (SCMs)

to systematically attribute responsibility in human-AI systems; measuring overall blameworthiness

using

while employing counterfactual reasoning to account for agents' expected epistemic levels. Two

Keywords: Responsibility Attribution, Causal Inference, Human-AI Collaboration; Decision-Making

case studies illustrate the framework 's adaptability in diverse human-AI collaboration scenarios.

1. Introduction

As Artificial Intelligence (Al) systems increasingly influence decision

-making

et al., 2021), the need to clearly define and attribute responsibility when outcomes are undesirable

Or

failures occur becomes crucial

complicates traditional accountability mechanisms due to the shared decision-making responsibili -

Santoni de Sio and Van den Hoven, 2018). The integration of AI

ties between humans and algorithms.

attribution; given their interactive, complex, and high-stakes nature

Human-AI systems pose unique challenges for responsibility

Amershi et al., 2019). On one

hand, the ability of humans to override or modify Al-driven decisions introduces an element of un-

predictability in outcomes. On the other hand, Al decisions often rely on huge datasets or complex

anticipate Al behaviour:

models that lack full transparency;

To address these

in multi-agent settings.

making it difficult for human counterparts to fully understand or

challenges; various methods have been proposed for responsibility attribution

Chockler and Halpern (2004)

and Halpern and Kleiman-Weiner (2018)

introduced frameworks to quantify blameworthiness based

was

On

causal relationships.

This definition

further extended to multi-agent settings by attributing blameworthiness through the Shapley

value, as demonstrated by

has been studied in decentralized , partially observable Markov decision processes

Friedenberg and Halpern (2019). Additionally; responsibility attribution

of actual causality (Triantafyllou et al., 2022) _

2016) to measure the degree of blameworthiness among agents.

Our code and data arc at

https:/ /github

Preprint. Under review.



0



0



63

2 weeks ago

@BPGPeixoto

4 Followers

Who Defines Me? A book about SJIA.

WHO

DEFINES

A

non-fictional book about

juvenile arthritis

These approaches rely on actual causality (Halpern,

com / yahang-qi/Causal-Attr-Human-AI.git.

the concept

using



BY

Beatriz Peixoto

1



1

@BPGPeixoto

A 10th grade project where I wrote a

book about SJIA

#personaljourney #sjiafoundation #chronicillness #personalproject

3 weeks ago



Home

Upload



0



67

1

1



Science Quiz

By Roshni Ranjan

0

0



mations (such

@yeeguy

applying

Golub:

Pu

ﶪ

Pus

+

+

+

+

+

blooming

Pu

2

M

FOLLOW



19

FOLLOW



13

FOLLOW



28

FOLLOW



FOLLOW



26

x

Use

28