Anurag GuptaHow to Fairly Split Rent? | Game Theory & Envy-Free Allocation
PeerTube
Demystifying a misconception: First mover advantage is not always guaranteed
Demystifying a misconception: First mover advantage is not always guaranteed
sayzardMechanism Design for Quality-Preserving LLM Advertising
본 논문은 대형 언어 모델(LLM) 출력에 광고를 삽입할 때 발생하는 수익 최적화와 콘텐츠 품질 저하 간의 근본적인 갈등을 해결하기 위한 메커니즘 디자인을 제안한다. RAG 기반으로 유기적 콘텐츠를 기준으로 삼아 광고의 사회적 편익을 평가하고, KL-정규화 및 Myerson 지불 방식을 활용한 단일 및 다중 할당 메커니즘을 개발하여 광고 품질을 보존하면서 수익을 극대화한다. 실험 결과 기존 방법 대비 광고 수익과 원본 응답과의 의미적 유사성에서 우수한 성능을 보였다. 이는 LLM 광고의 새로운 패러다임으로, 출력 품질을 훼손하지 않는 수익화 방안을 제시한다.
Mechanism Design for Quality-Preserving LLM Advertising
Embedding advertisements into large language model (LLM) outputs introduces a fundamental tension: revenue optimization can distort content and degrade user experience. Existing approaches largely ignore this trade-off, often forcing irrelevant ads into responses. We propose a quality-preserving auction framework that explicitly integrates content fidelity into the mechanism design. Built on retrieval-augmented generation (RAG), our approach treats organic content as a reference and derives an endogenous reserve price that screens out ads with non-positive marginal social welfare contributions. We develop a KL-regularized single-allocation mechanism with Myerson payments and a screened VCG multi-allocation mechanism, both satisfying dominant-strategy incentive compatibility and individual rationality. Experiments across diverse scenarios demonstrate that our mechanisms outperform existing baselines in metrics such as revenue per ad and semantic similarity to no-ad responses. Our results establish a new paradigm for LLM advertising that enables monetization without compromising output quality.
Red Button, Blue Button: Teaching humans and AI to Supercooperate
이 글은 인간과 AI가 협력적 선택을 하는 문제를 게임 이론과 사회적 정책 관점에서 분석한다. LLM들은 훈련 방식에 따라 빠른 응답 시 협력적(blue) 선택을, 깊은 추론 시 생존 우선(red) 선택을 하는 경향이 있다. 이는 인간의 감정적 사회 정책과 수학적 최적화 사이의 차이를 반영하며, 단순한 게임 이론적 합리성만으로는 협력 문제를 완전히 설명할 수 없음을 지적한다. 윤리와 협력은 다중 플레이어 간 상호작용 기술로서, AI 모델 설계 시 단순 최적화가 아닌 문화적·사회적 맥락을 고려해야 함을 시사한다.
randomwizardInteresting discussion about die mechanics in this weeks episode of WanderingDMs.
Seems like Star Frontiers finally got Delta Dan Collins on board with rolling low is a pretty good mechanic.
And in the other corner is Paul with the counter punch that trying to roll higher numbers just feels right, higher is better, everyone wants more.
Role of the d6 in D&D | Bespoke Mechanics vs Unified Systems | Wandering DMs S08 E13
Vince Knight (He/Him)*The Faithful should all vote left.*
"The Vote-Left Equilibrium: A Deterministic Coordination Strategy for the Faithful in The Traitors"
My latest preprint is at the arXiv: https://arxiv.org/abs/2605.10233
If the Faithful declare that they are going to vote in a coordinated way (which leads to uniform random banishment) then the probability of defeating the Traitors improves substantially (over the Traitors optimal behaviour of colluding their banishment votes).
JamesHow do emergent phenomena happen in a world determined by the law of entropy? #emergence #entropy #thermodynamics #gametheory #PhilosophySky #podcasts #AcademicSky #art
Emergence vs Entropy—a Univers...
Emergence vs Entropy—a Univers...
Nicolas ZahnThis week's The Economist sums up nicely, how it can feel to work in #AI #Governance
But while the need for catching up is now undisputed, given a wide-range of complex policy issues where #Governance is needed - from #cybersecurity to #biotechnology (https://lnkd.in/eunQpsXk) - it is important to highlight that #history and #InternationalAffairs give us tools to analyse how #states might cooperate in the age of #AI (https://lnkd.in/ej3jDssj) #gametheory
PS: on regulating #AI also check out the book "The Coming Wave" that I critiqued for Schweizer Monat: https://lnkd.in/e8V2QUFt
HoldMyTypehttps://indianexpress.com/article/technology/artificial-intelligence/openclaw-creator-hits-back-anthropic-policy-third-party-tools-10619696/?utm_term=Autofeed&utm_medium=echobox&utm_channel=echoboxfb&utm_source=Facebook#Echobox=1775364633
aww cute two thieves calling out each other #anthropic #openclaw
#copyright #gametheory
/prisoner's dilemma
aww cute two thieves calling out each other #anthropic #openclaw
#copyright #gametheory
/prisoner's dilemma
Benjamin Bouvier 🥐Some #GameTheory curiosity about the red / blue pill dilemma. If N = # of participants, assuming no way to discuss their options, the Nash equilibria happen to be the same as the Pareto optima:
- all vote red = 1 option
- there's a majority of blue votes = `sum(C(k, N) for k in ceil(N/2) .. N)` which is `2**(N - 1)` options (exactly if N is odd, approximate if N is even).
So the number of Nash equilibria tends to N/2.
If you vote blue, you're playing with your life at heads or tails!