Earlier this month I was in Seoul for ICML 2026, presenting our paper When Agents Lie at the NExT-Game workshop. Between sessions, I did what I always do at conferences: wander the poster halls until my feet gave out.

With thousands of papers at ICML, you need a filter. Mine is pretty simple:

Here's what made the cut, in no particular order. For each paper: a quick overview, and why I stopped at the poster.

Discovering Ordinary Differential Equations with LLM-Based Qualitative and Quantitative Evaluation

Strategic Exploitation in LLM Agent Markets: A Simulation Framework for E-Commerce Trust

Solipsistic Superintelligence is Unlikely to be Cooperative

Cooperate to Compete: Strategic Coordination in Multi-Agent Conquest

Computational Arbitrage in AI Model Markets

Systematic Failures in Collective Reasoning under Distributed Information in Multi-Agent LLMs

BrokenMath: A Benchmark for Sycophancy in Theorem Proving with LLMs

PaperBanana: Automating Academic Illustration for AI Scientists

Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge

Kantian Equilibrium in the Age of Multi-Agent Systems

Benchmarking at the Edge of Comprehension


That's the list. If you were at ICML and caught a poster I missed that fits the "readable, applicable, one clean sentence" bar, send it my way at JerickS.1380@gmail.com.