AutoTavern: Self-Improving LLM Agents Build a Hearthstone AI Rivaling Top Players
Changlian Ma · hf · 2026-10-27
Researchers from Nanjing University and Shanghai AI Laboratory present AutoTavern, showing that self-improving LLM research agents can autonomously carry a hard real-world problem from raw data to a system that holds its own against people.
Key points:
- Target domain: Hearthstone Battlegrounds, an eight-player auto-battler with stochastic shops, hidden information, and nearly a thousand interacting cards.
- Specialized LLM research agents share a typed research memory; every claim passes verification gates, and live games close the feedback loop to improve the simulator, opponents, and policy.
- Humans only set goals and adjudicate ambiguous mechanics; the resulting RL agent surpasses top human players.
More from AGI Musings
- Commentary: OpenAI won't release reasoning traces because the reasoning process is the real product — max_paperclips · 2026-10-27
- Reddit user argues cheap decentralized AI is the only escape from compute concentration — CarzyForTech · 2026-10-27
- Anthropic internal model independently refutes Exponential Time Hypothesis, found Oct 16 — beffjezos · 2026-10-27
- Anthropic at $2T valuation with ~5,000 staff vs ZIRP-era SaaS giants, in one chart — beffjezos · 2026-10-27
- Physicist on OpenAI's math result: it's mathematical physics, not theoretical physics — skdh · 2026-10-27
- Applied AI companies are playing katamari: get big, then bet on domain capabilities — saranormous · 2026-10-27