一篇探讨对比学习框架下弱到强泛化的新论文，理论分析和实验验证结合，为AI大模型泛化研究提供新视角。
arXiv:2510.07884v2 Announce Type: replace-cross Abstract: Weak-to-strong generalization provides a promising paradigm for scaling large language model…
颠覆认知：GNN本质是高级启发式算法而非特征学习器，理论推导+实验验证给出新视角
arXiv:2601.13465v4 Announce Type: replace Abstract: Graph neural networks are usually treated as auxiliaries for combinatorial optimization: they imit…
理论揭示：判别模型的条件分布等价性如何约束内部表示的唯一性，为理解模型表征相似性提供新视角
arXiv:2602.15438v3 Announce Type: replace-cross Abstract: For a broad family of discriminative models that includes autoregressive language models, id…
探讨大语言模型与人类表征模式的差异，从认知科学视角剖析AI理解能力
arXiv:2606.21616v1 Announce Type: new Abstract: Much work on the cognitive foundations of AI has focussed on comparisons between the ways in which Lar…
探讨LLM作为语言学模态模型的新视角，论文从理论层面解析大语言模型的语言学应用价值
arXiv:2606.10467v1 Announce Type: new Abstract: The rapid advancement of large language models (LLMs) has intensified debates about their significance…
深入剖析注意力机制中标准化的局限性，揭示模型选择性能力与token选择的几何分离理论边界。
arXiv:2508.17821v3 Announce Type: replace-cross Abstract: This paper investigates the limitations of the normalization in attention mechanisms. We beg…
探讨基于分数的生成模型在持续学习中的遗忘与稳定性问题，为提升模型鲁棒性提供新视角。
arXiv:2601.21868v2 Announce Type: replace-cross Abstract: Understanding the stability and long-time behavior of generative models is a fundamental pro…
颠覆深度缩放传统认知：层间相似性导致“越深越差”的逆缩放现象，ICML 2026最新发现。
arXiv:2602.05970v2 Announce Type: replace Abstract: Neural scaling laws relate loss to model size in large language models (LLMs), yet depth and width…
揭秘线性循环记忆在部分可观测强化学习中的有效性，带来简洁而深刻的理论解释
arXiv:2605.31261v1 Announce Type: cross Abstract: The family of linear recurrent neural networks has shown strong performance as recurrent memory unit…
从理论不可能到局部补丁，这篇论文系统解析了LLM错误架构，为提升模型可靠性提供全新视角。
arXiv:2605.30628v1 Announce Type: cross Abstract: Universal LLM reliability is not a finite-library problem: across all possible tasks, tools, schemas…
ICML '26论文揭示LLM思维链token复杂度的BAPO边界，理论深度与实证兼顾。
arXiv:2602.02909v2 Announce Type: replace Abstract: Inference-time scaling via chain-of-thought (CoT) reasoning is a major driver of state-of-the-art …
揭秘LoRA微调大模型时如何“记住”信息，提出参数化记忆定律，深入理论分析注意力机制。
arXiv:2605.30260v1 Announce Type: cross Abstract: Large Language Models (LLMs) must continuously learn and update knowledge to remain effective in dyn…
从有限样本视角揭示扩散后验采样器的失效机制与深层原因
arXiv:2605.30330v1 Announce Type: new Abstract: Diffusion models have excellent capacity to model complex distributions of natural data, which has mad…
黑盒环境下LLM水印逃避新突破：理论揭示重写型攻击无需查询即可规避检测，且不损语义完整性，为AI内容溯源安全敲响警钟。
arXiv:2509.23019v5 Announce Type: replace-cross Abstract: Watermarking offers a promising solution for detecting LLM-generated content, yet its robust…
Kalai和Vempala提出的概率框架，为大语言模型幻觉现象提供了严谨的形式化定义与校准条件，是理论突破。
arXiv:2605.26808v1 Announce Type: cross Abstract: Hallucination is a central limitation of large language models (LLMs), and substantial effort has be…
从动力系统与熵相变视角揭示LLM推理的触发条件，为理解大模型认知机制提供新洞见。
arXiv:2605.22873v1 Announce Type: cross Abstract: Chain-of-thought (CoT) reasoning has become the default strategy for enhancing LLM capabilities, yet…
探索预测编码网络在无限宽度与深度极限下的理论性质，揭示其动态行为与收敛性，为深度学习提供新视角。
arXiv:2602.07697v2 Announce Type: replace-cross Abstract: Predictive coding (PC) is a biologically plausible alternative to standard backpropagation (…
VAE新视角：用交换子数学工具量化模型不确定性，理论严谨有深度
arXiv:2605.23449v1 Announce Type: new Abstract: Variational autoencoders (VAEs) often struggle to represent non-commutative structure in learned laten…
揭示自我对弈仅在自合成数据提供可学习信息增益时才有效演化，为AI训练策略提供关键理论指导。
arXiv:2603.02218v2 Announce Type: replace Abstract: Large language models (LLMs) make it plausible to build systems that improve through self-evolving…
从贝叶斯几何视角重新阐释Transformer注意力机制，揭示其内在概率结构。
arXiv:2512.22471v5 Announce Type: replace Abstract: Transformers often appear to perform Bayesian reasoning in context, but verifying this rigorously …