Memory is Key for Long-Horizon Agents: 15 LLMs Manage a Football Club for 20 Years
dair_ai · x · 2026-08-30
A study tested long-horizon memory and planning by having LLM agents manage a football club for 20 in-game seasons (approx. 340-400 decision stops).
Key Findings:
- All 15 frontier models survived the entire horizon, while most scripted baselines failed.
- Scale, price, vendor, or token spend did not predict rankings; order only settled late in the run.
- Top models exhibited specific managerial behaviors: cutting slow-payoff investments near the end and opening contract renewals well before deadlines.
- Two universal failures were found: no model learned the market's hidden prices from rejected bids, and self-managed memory also failed.
The research highlights that memory/recall is critical for breaking long-horizon agent bottlenecks.
More from Research
- Cohere Labs world model course session 4: DreamerV1-V4 architectures in one go — Cohere_Labs · 2026-10-16
- Reasonable post-trains Nemotron 3.5 on 3,000 Verus proofs, lifting pass@1 ~30 points to 73.3% — fhuszar · 2026-10-16
- GPT-6.1 Sol audit finds 6% of arXiv papers had false main theorems, down to 1.5% recently — nitarshan · 2026-10-16
- One tiny word changes the conversation: how LLMs pick up on 'just' — StellaLisy · 2026-10-16
- SEAR: first large-scale controlled study of self-evolving robot agents, 7 LLMs, 330 tasks — xiye_nlp · 2026-10-16
- After Claude-assisted elliptic curve records fell, EC Rank challenge opens 156 record slots — Jsevillamol · 2026-10-16