1
Enhancing LLM Safety Through a Theoretical Minimax Game Lens
从博弈论角度破解大模型安全难题，提出理论新框架。
arXiv:2502.05163v2 Announce Type: replace-cross Abstract: The rapid advancement of large language models (LLMs) necessitates effective mechanisms to e…
2
Green AI: A Unified Theory of Computational Waste
从根源量化AI计算浪费，提出绿色AI统一理论框架，减少环境成本。
Article URL: https://zenodo.org/records/20459312 Comments URL: https://news.ycombinator.com/item?id=48411690 Points: 1 # Comments: 0
3
Greed is Good: A Unifying Perspective on Guided Generation
从“贪婪”视角统一引导式生成，为AI生成方法提供全新理论框架，已被NeurIPS 2025接收。
arXiv:2502.08006v3 Announce Type: replace-cross Abstract: Training-free guided generation is a widely used and powerful technique that allows the end …
4
A Task-Centric Theory for Iterative Self-Improvement with Easy-to-Hard Curricula
从易到难课程驱动模型自我迭代，任务中心理论揭示高效训练新路径。
arXiv:2602.10014v3 Announce Type: replace Abstract: Iterative self-improvement fine-tunes an autoregressive large language model (LLM) on reward-verif…
5
A Theoretical Framework for Self-Play Theorem Proving Algorithms
自对弈算法在定理证明领域有了严谨的理论根基，为AI数学推理提供新视角。
arXiv:2606.01861v1 Announce Type: new Abstract: Self-play, a type of training algorithm that enables a model to self-improve, has recently shown promi…
6
Beyond the Bellman Recursion: A Pontryagin-Guided Framework for Non-Exponential Discounting
用庞特里亚金最优控制原理突破传统强化学习折扣框架，为认知与经济学中的非指数折扣提供新解法。
arXiv:2605.20996v1 Announce Type: new Abstract: Most value-based and actor--critic reinforcement learning methods rely on Bellman-style recursions, ye…
7
Consciousness as Uncommon Self-Knowledge: A Synergistic Information Framework
从信息分解视角重新定义意识：提出“非共同自我知识”作为意识候选标准，摘要揭示协同信息的核心作用。
arXiv:2605.13884v1 Announce Type: cross Abstract: We propose uncommon self-knowledge (USK) as a candidate criterion for consciousness: synergistic inf…