1
OpenAI 新董事严厉警告：AI 失控将是灾难性的，大多数人可能丧命
北京时间 9 月 10 日，据《商业内幕》报道，OpenAI 周三宣布保罗 · 克里斯蒂亚诺 (Paul Christiano) 加入其基金会董事会。随后，克里斯蒂亚诺直言不讳地谈到了 AI 超级智能时代所面临的风险。 克里斯蒂亚诺此前曾在 OpenAI 负责对齐研究，后来进入美国联邦政府工作，担任…
2
Geometry of Forgetting: Representation Flux in Continual Learning
从几何视角剖析持续学习中的表征漂移，揭示灾难性遗忘的内在机制，为构建不退化的智能模型提供新思路。
arXiv:2608.15854v1 Announce Type: new Abstract: Catastrophic forgetting remains a fundamental obstacle to continual learning, where neural networks lo…
3
(How) Learning Rates Regulate Catastrophic Overtraining
学习率如何抑制灾难性过训练？这项研究揭示关键调节机制，为深度学习训练提供全新视角。
arXiv:2604.13627v2 Announce Type: replace Abstract: Supervised fine-tuning (SFT) is a common first stage of LLM post-training, teaching the model to f…
4
能无限制生成“灾难”图片，谷歌地球 AI 生图功能上线不到一天就下架
输入文字指令，就能把谷歌地球影像改造成任意场景，AI 生图潜力惊人，可惜上线一天就下架了。
IT之家 8 月 1 日消息，谷歌为 Google Earth（谷歌地球）加入 AI 图像编辑功能后， 不到 24 小时便紧急下线 。这一功能调用 Gemini 旗下 Nano Banana 2 图像模型，用户只需输入文字指令，便可修改谷歌地球中的地点影像。 该功能上线后很快遭到滥用。AI 研究人员…
5
亿万富翁想把数据中心“送上天”，科学家警告将造成生态灾难
IT之家 7 月 29 日消息，过去一年，马斯克、杰夫 · 贝索斯等亿万富翁逐渐迷上了难以置信的构想：既然地面社区不欢迎数据中心，与其强行推动建设，不如把它们全部送上太空。 多名专家指出，无论成本还是技术和运营条件，太空数据中心都“近乎荒诞”。即便企业真能克服实际困难，在地球轨道大规模部署此类设施，…
6
Mixtures of SubExperts for Large Language Continual Learning
提出混合子专家架构，让大模型持续学习时有效缓解灾难性遗忘，兼顾精度与效率。
arXiv:2511.06237v2 Announce Type: replace Abstract: Enabling lifelong learning in LLMs demands resolving the stability-plasticity dilemma (i.e., model…
7
ReCoLoRA: Spectrum-Aware Recursive Consolidation for Continual LLM Fine-Tuning
突破LLM持续微调遗忘瓶颈，ReCoLoRA以频谱感知递归合并实现任务序列高效学习。
arXiv:2607.07719v1 Announce Type: new Abstract: Parameter-efficient fine-tuning adapts a large language model to one task cheaply, but across a task s…
8
Learning Probabilistic Prompt for Continual Learning
用概率提示应对持续学习中的灾难性遗忘，ECCV 2026 新方法值得关注
arXiv:2607.04711v1 Announce Type: new Abstract: Continual learning aims to progressively learn from a sequence of tasks, each containing a disjoint su…
9
The Bayesian Approach to Continual Learning: An Overview
持续学习如何对抗遗忘？贝叶斯方法综述给你严谨答案，从核心原理到前沿进展一网打尽。
arXiv:2507.08922v3 Announce Type: replace-cross Abstract: Continual learning is an online paradigm where a learner continually accumulates knowledge f…
10
Hidden Forgetting in Continual Multimodal Learning: When Accuracy Survives but Grounding Fails
多模态持续学习藏有“隐性遗忘”：模型精度不变，但跨模态对齐已悄悄失效，值得关注。
arXiv:2607.02020v1 Announce Type: new Abstract: Multimodal large language models must continually adapt to evolving tasks and domains, yet standard co…
11
Curvature-Guided Mixing for MLLM Adaptation
理论严谨的曲率引导混合方法，有效缓解多模态大模型微调中的灾难性遗忘。
arXiv:2606.24963v1 Announce Type: cross Abstract: Fine-tuning Multimodal Large Language Models (MLLMs) on specialized tasks often leads to catastrophi…
12
Representation Interventions Enable Lifelong Knowledge Memory Control in LLMs
通过表示层干预让大语言模型终身可控记忆知识，不遗忘旧信息，大模型记忆管理新范式。
arXiv:2511.20892v4 Announce Type: replace Abstract: Large language models (LLMs) often produce incorrect or outdated content after being employed. Eff…
13
Forged Calamity: Benchmark for Cross-Domain Synthetic Disaster Detection in the Age of Diffusion
扩散模型生成逼真的合成灾难图像，首个跨域检测基准如何应对伪造危机？
arXiv:2606.18554v1 Announce Type: new Abstract: The rapid advancement of text-to-image diffusion models has enabled the creation of highly photorealis…
14
Continual Learning with Support Boundary Experience Blending
新方法「支持边界经验混合」解决持续学习中的灾难性遗忘，理论创新与实践效果兼具。
arXiv:2507.23534v3 Announce Type: replace Abstract: Continual learning (CL) seeks to mitigate catastrophic forgetting when models are trained with seq…
15
FOGO: Forgetting-aware Orthogonalization Optimizer
新型FOGO优化器通过遗忘感知正交化机制，有效缓解持续学习中的灾难性遗忘问题
arXiv:2606.10406v1 Announce Type: cross Abstract: We argue that forgetting is not confined to continual learning but is a general optimization phenome…
16
You can't detect your way out of catastrophic LLM failure
检测技术无法根除大模型灾难性失败，一场揭示模型能力边界的真实辩论。
Article URL: https://github.com/joseteiadirector/teia-igo-vs-claude-opus-4.8/blob/main/README.en.md Comments URL: https://news.ycombinator.com/item?id…
17
CRMA: A Spectrally-Bounded Backbone for Modular Continual Fine-Tuning of LLMs
用频谱有界骨干解决大模型持续微调中灾难性遗忘与跨任务改进的矛盾，模块化方案有新思路
arXiv:2606.00382v1 Announce Type: new Abstract: Sequential fine-tuning of large language models forces a choice: let the shared substrate keep learnin…
18
Overcoming Forgetting in LLM Fine-Tuning with Evolution Strategies
用进化策略对抗微调遗忘，提升LLM持续学习能力的新方法
arXiv:2605.30148v1 Announce Type: cross Abstract: Evolution Strategies (ES) has recently emerged as a competitive alternative to reinforcement learnin…
19
Here's why the failure of Blue Origin's New Glenn rocket is so catastrophic
Blue Origin New Glenn火箭首秀失利，细数为何这次失败打击远超想象
"I hope that it makes it far enough away from the pad that it does not cause pad damage."
20
On-Policy Replay for Continual Supervised Fine-Tuning
提出On-Policy Replay方法，解决持续监督微调中的灾难性遗忘问题，为LLM高效增量学习提供新思路。
arXiv:2605.29495v1 Announce Type: new Abstract: Continual supervised fine-tuning (SFT) is the de facto recipe for adapting large language models (LLMs…