iiinova — where your digital self keeps evolving

Vinzenz Thoma @vthoma.bsky.social · 07/04/2026

Must be good then, if you are already in chapter 5 & learning new things... Can't wait!

010

Vinzenz Thoma @vthoma.bsky.social · 06/04/2026

Bought this as well now & looking forward to read. Thanks for sharing!

110

Vinzenz Thoma @vthoma.bsky.social · 21/03/2026

Thank you, appreciated! Concerning opens-source: We discussed this and would all be happy to review an implementation for openspiel.

120

Vinzenz Thoma @vthoma.bsky.social · 16/03/2026

6/6 🧵Future Work: We hope deep incentive design can serve as a general-purpose tool for people to build on. If you have an incentive design problem, plug in your loss/problem instance or feel free to reach out!

130

Vinzenz Thoma @vthoma.bsky.social · 16/03/2026

5/6 🧵Results: We validate on three tasks: multi-agent contract design, machine scheduling, and inverse equilibrium problems. For each, a single network handles the *full* distribution of problem instances across all game sizes from 2×2 to 16×16.

130

Vinzenz Thoma @vthoma.bsky.social · 16/03/2026

4/6 🧵The framework (see figure): We learn the (unique) equilibrium function with a pretrained "differentiable equilibrium block" and backpropagate through it to train our mechanism generator on the whole distribution of problems—no per-instance optimization at test time.

130

Vinzenz Thoma @vthoma.bsky.social · 16/03/2026

3/6 🧵The idea: Using max-entropy (coarse) correlated equilibria renders the bilevel problem differentiable. Thereby we unlock the whole toolkit of machine learning and gradient-based optimization to tackle this game-theoretic problem.

130

Vinzenz Thoma @vthoma.bsky.social · 16/03/2026

2/6 🧵The problem: You're a designer who (partially) controls the rules of a game and agents in response play an equilibrium. How do you set the rules so the resulting behavior aligns with your objective? This is incentive design and it shows up in contract & mechanism design, machine scheduling etc.

150

Vinzenz Thoma @vthoma.bsky.social · 16/03/2026

[1/6] 🧵Hi there! Our paper "Deep Incentive Design with Differentiable Equilibrium Blocks" is out now, born from my internship at Google DeepMind with @lukemarris.bsky.social and Georgios Piliouras. Thread below! Paper: arxiv.org/abs/2603.07705

1234

Reposted by Vinzenz Thoma

drimgemp.bsky.social @drimgemp.bsky.social · 27/01/2026

If ICLR is any indication, LLMs + Game Theory / Multi-Agent is thriving. We'd love to see your research ideas at AAMAS this May in Cyprus! Submission deadline is Feb 4. More details below.

0154

Vinzenz Thoma @vthoma.bsky.social · 22/12/2025

@sharky6000.bsky.social , we already did:) See here: bsky.app/profile/vtho...

110

Vinzenz Thoma @vthoma.bsky.social · 18/12/2025

Unlike board games, real-world strategic interactions are messy. Traditional game theory thus needs a boost for the age of agentic AI. Our #AAMAS2026 workshop "Strategic Engineering"(sites.google.com/view/se-aama...) in Cyprus aims to bridge the gap. Come join us to unlock truly strategic AI!

0136

Vinzenz Thoma @vthoma.bsky.social · 22/04/2025

[1/2]Our work “Learning to steer Markovian Agents under Model Uncertainty” explores the problem of steering (unknown) learning dynamics in Markov Games towards desirable outcomes (e.g. Pareto optimal Nash). My co-author Jiawei Huang is presenting it at ICLR on Apr 24, 3:00-5:30pm GMT+8, Poster #401.

242

Reposted by Vinzenz Thoma

Robin Ranjit Singh Chauhan @robinchauhan.bsky.social · 10/03/2025

E65: NeurIPS 2024 – Posters and Hallways 3 - Claire Bizon Monroc of Inria : WFCRL for Wind Farm Control Andrew Wagenmaker of @ucberkeleyofficial.bsky.social : Leveraging Simulation to Bridge Sim-to-Real Gap - @harwiltz.bsky.social of @mila-quebec.bsky.social : Multivariate Distributional RL (cont)

132

Vinzenz Thoma @vthoma.bsky.social · 26/02/2025

I'm at AAAI this week, presenting our paper “Computing Perfect Bayesian Equilibria in Sequential Auctions” (arxiv.org/abs/2312.04516) done at @eth-ai-center.bsky.social. If interested, join the poster session (#657 on Saturday 12:30 – 14:30) or oral presentation (room 115C on Sunday 2pm).

arxiv.org

Computing Perfect Bayesian Equilibria in Sequential Auctions with Verification

We present an algorithm for computing pure-strategy epsilon-perfect Bayesian equilibria in sequential auctions with continuous action and value spaces. Importantly, our algorithm includes a verificati...

071