<![CDATA[c_cpp_csharp的专栏]]>https://blog.csdn.net/c_cpp_csharpzh-cnhttps://blog.csdn.net/<![CDATA[Component and Dimension Sparsity in Transformer Refusal Mechanisms 论文总结]]>https://blog.csdn.net/c_cpp_csharp/article/details/168641211https://blog.csdn.net/c_cpp_csharp/article/details/168641211c_cpp_csharpWed, 28 Oct 2026 12:00:00 +0800 论文信息：arXiv:2610.06903，接收于COLM 2026，作者：Vincen]]><![CDATA[CodeGraph: Open‑Taxonomy Knowledge Graph for Source Code with Wikidata Grounding]]>https://blog.csdn.net/c_cpp_csharp/article/details/168641368https://blog.csdn.net/c_cpp_csharp/article/details/168641368c_cpp_csharpWed, 28 Oct 2026 09:00:00 +0800 论文arXiv链接：[https://arxiv.org/abs/26]]><![CDATA[LogicTrack: Auditing Reasoning Trajectories of Large Language Models with Formal Logic Solvers 论文总结与]]>https://blog.csdn.net/c_cpp_csharp/article/details/168643755https://blog.csdn.net/c_cpp_csharp/article/details/168643755c_cpp_csharpWed, 28 Oct 2026 08:00:00 +0800 论文arXiv链接：htt]]><![CDATA[Accelerating Dense LLMs via L0‑regularized Mixture‑of‑Experts 论文总结与翻译]]>https://blog.csdn.net/c_cpp_csharp/article/details/168643895https://blog.csdn.net/c_cpp_csharp/article/details/168643895c_cpp_csharpWed, 28 Oct 2026 07:00:00 +0800 论文arXiv链接：https://arxiv.org/abs/2609.21672 ##]]><![CDATA[Abstention and Noise Filtering: Two Missing Primitives of Softmax Attention 论文阅读总结]]>https://blog.csdn.net/c_cpp_csharp/article/details/168643792https://blog.csdn.net/c_cpp_csharp/article/details/168643792c_cpp_csharpTue, 27 Oct 2026 13:00:00 +0800 论文arXiv链接：[https://arxiv.org/abs/]]><![CDATA[Block‑Sparse Attention with Semantic‑Geometric Decoupled Routing（arXiv:2609.22884）论文总结与翻译]]>https://blog.csdn.net/c_cpp_csharp/article/details/168643846https://blog.csdn.net/c_cpp_csharp/article/details/168643846c_cpp_csharpTue, 27 Oct 2026 12:00:00 +0800 论文链接：[https://arxiv.org/ab]]><![CDATA[GVPO++: Group Variance Policy Optimization for LLM Post‑Training and On‑Policy Distillation]]>https://blog.csdn.net/c_cpp_csharp/article/details/168643973https://blog.csdn.net/c_cpp_csharp/article/details/168643973c_cpp_csharpTue, 27 Oct 2026 10:00:00 +0800 论文链接：https://arxiv.org/a]]><![CDATA[RBS‑Attention 论文阅读总结]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644008https://blog.csdn.net/c_cpp_csharp/article/details/168644008c_cpp_csharpTue, 27 Oct 2026 09:00:00 +0800 论文信息：arXiv:2609.20971，RBS‑Attention: Radius‑Bounded Sparse Prefill for Long‑Context Large Langu]]><![CDATA[Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning（arXiv:2609.03430）]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644109https://blog.csdn.net/c_cpp_csharp/article/details/168644109c_cpp_csharpTue, 27 Oct 2026 08:00:00 +0800<![CDATA[SWE‑Gate 论文阅读总结]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644136https://blog.csdn.net/c_cpp_csharp/article/details/168644136c_cpp_csharpTue, 27 Oct 2026 07:00:00 +0800 论文信息：arXiv:2609.04167，SWE‑Gate: Passing Functional Tests Is Not Enough for Software Engineering Agen]]><![CDATA[What Matters for Aggressive Decoding‑Time KV Eviction? Temporal Aggregation and Ranking Preservation]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644042https://blog.csdn.net/c_cpp_csharp/article/details/168644042c_cpp_csharpMon, 26 Oct 2026 15:00:00 +0800 arXiv:2609.0351]]><![CDATA[GrowPage: On‑Demand KV Budgeting for Efficient LLM Reasoning Serving（arXiv:2609.03494）]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644078https://blog.csdn.net/c_cpp_csharp/article/details/168644078c_cpp_csharpMon, 26 Oct 2026 14:00:00 +0800 论文链接：[https://arxiv.org/abs/2]]><![CDATA[Large Language Models in Resolving Contextual Knowledge Conflicts（arXiv:2609.03148）]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644162https://blog.csdn.net/c_cpp_csharp/article/details/168644162c_cpp_csharpMon, 26 Oct 2026 13:00:00 +0800 会议：EMNLP 2026 > 论文链接：https://arx]]><![CDATA[RuleMem: Active Rule Memory for Long‑Term Conversational Agents 论文总结与翻译]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644192https://blog.csdn.net/c_cpp_csharp/article/details/168644192c_cpp_csharpMon, 26 Oct 2026 12:00:00 +0800 论文arXiv链接：https://arxiv.org/abs/2609.03915 ]]><![CDATA[Representational alignment yields generalizable safety in language models]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644279https://blog.csdn.net/c_cpp_csharp/article/details/168644279c_cpp_csharpMon, 26 Oct 2026 11:00:00 +0800 说明：arXiv网页摘要接口仅公开**Abstract**，**Introduction、Conclusion**正文内容不在返回的元数据中，无法直接获取原文，因此仅完成摘要原文+翻译；文章主要内容、创新点基于摘要信息整理。 # Re]]><![CDATA[It‘s the Problem, Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectories]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644313https://blog.csdn.net/c_cpp_csharp/article/details/168644313c_cpp_csharpMon, 26 Oct 2026 10:00:00 +0800 论文arXiv地址：https://arxi]]><![CDATA[论文总结：Doesn‘t Stop Reasoning: Analysis of Spurious CoT Termination]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644491https://blog.csdn.net/c_cpp_csharp/article/details/168644491c_cpp_csharpMon, 26 Oct 2026 09:00:00 +0800 论文arXiv：https://arxiv.org/abs/2609.03633 > 会议：EMNL]]><![CDATA[GeoEdit: Geometric Knowledge Editing for Large Language Models]]>https://blog.csdn.net/c_cpp_csharp/article/details/168539988https://blog.csdn.net/c_cpp_csharp/article/details/168539988c_cpp_csharpMon, 26 Oct 2026 08:00:00 +0800<![CDATA[Survey of Large Multimodal Reasoning Models (LMRMs): From Modular to Native (arXiv:2505.04921)]]>https://blog.csdn.net/c_cpp_csharp/article/details/168460583https://blog.csdn.net/c_cpp_csharp/article/details/168460583c_cpp_csharpMon, 26 Oct 2026 07:00:00 +0800<![CDATA[MemoryLACE论文总结与翻译]]>https://blog.csdn.net/c_cpp_csharp/article/details/168644242https://blog.csdn.net/c_cpp_csharp/article/details/168644242c_cpp_csharpSun, 25 Oct 2026 16:00:00 +0800 论文信息：arXiv:2609.03201，MemoryLACE: Memory Lifecycle‑Aware Consolidation and Evidence Retrieval ## ]]>