Memento 3：冻结 LLM 通关全部 25 个 ARC-AGI-3 游戏
Memento 3 不更新 LLM，靠自写规则库通关 25 个 ARC-AGI-3 游戏，做法很巧。
论文提出 Memento 3，在底层 LLM 完全冻结、不更新参数的情况下通关 ARC-AGI-3 全部 25 个公开游戏的所有关卡，动作数只有人类的 44%。Memento 3 让智能体把对环境的理解写成自然语言规则库，再编译成预测下一状态的代码，新版本只有在重放时逐格复现全部已观察的状态转移才会保留。在 Atari Pong 案例中，这套自学的预测模型驱动控制器以 21 比 0 赢下三局，中途不再调用 LLM。
Insightful paper on recursive self-improvement with a frozen LLM.
(bookmark it)
Memento 3 clears every level of all 25 public ARC-AGI-3 games and uses 44% of the human action count. The underlying LLM is never updated.
How it works:
The agent writes what it believes about the environment as a natural-language rulebook, then compiles that rulebook into code that predicts the next state. A new version is kept only if replaying the code reproduces every transition the agent has observed, cell for cell.
In an Atari Pong case study, the learned model drives a controller that wins 21 to 0 in three games without further LLM calls.
https://t.co/aB2hr3PXtZ
Chat with Paper: https://t.co/dtCxhYAh8L
