面向设备端LLM推理的KV缓存碎片引擎，跳过最昂贵推理环节，专为ARM64 Android优化。
Article URL: https://github.com/bossandboss/EdgeSync-LLM Comments URL: https://news.ycombinator.com/item?id=48732973 Points: 2 # Comments: 0
语音大模型解码器层存在多少冗余？这篇论文提出测量方法，为模型压缩提供新视角
arXiv:2603.05121v2 Announce Type: replace-cross Abstract: Speech Large Language Models route speech encoder representations into an LLM decoder that t…
OpenAI与博通联合打造3nm AI推理芯片Jalapeño，年底部署，第二代更将导入A16节点。
IT之家 6 月 26 日消息，OpenAI 与 Broadcom（博通）当地时间本月 24 日宣布了两家企业合作开发的 LLM 优化 AI 推理 ASIC 芯片 Jalapeño，目标在今年底实现初步部署。 台媒《经济日报》今日援引供应链消息报道称， Jalapeño 芯片基于 3nm 工艺制程 …
OpenAI联手博通推出专为大模型推理优化的芯片，硬件加速或将改写算力格局。
Article URL: https://openai.com/index/openai-broadcom-jalapeno-inference-chip/ Comments URL: https://news.ycombinator.com/item?id=48659257 Points: 11 …
无需额外LLM，智能路由你的提示词——Wayfinder通过轻量级规则降低大模型调用成本。
Article URL: https://github.com/itsthelore/wayfinder-router Comments URL: https://news.ycombinator.com/item?id=48648150 Points: 1 # Comments: 0
轻量级潜推理技术利用强化学习优化LLM生成推理轨迹，提升叙事类任务表现
arXiv:2512.02240v2 Announce Type: replace Abstract: Large language models (LLMs) tackle complex tasks by generating long chains of thought or "reasoni…
企业级LLM优化新利器OptiKIT，自动满足SLO并大幅缩短优化时间，MLSys 2026收录论文。
arXiv:2601.20408v2 Announce Type: replace-cross Abstract: Enterprise LLM deployment faces a critical scalability challenge: organizations must optimiz…
大模型训练新突破：灵活上下文并行策略，高效扩展LLM训练边界
arXiv:2602.21788v2 Announce Type: replace-cross Abstract: Scaling long-context capabilities is crucial for Large Language Models (LLMs). However, real…
无人机低空经济网络如何用LLM优化机载视觉语言推理？新方案实现高效多模态实时推断。
arXiv:2510.10028v2 Announce Type: replace Abstract: The rapid advancement of Low-Altitude Economy Networks (LAENets) has enabled a variety of applicat…
仅靠排序即可实现大模型张量化，比传统分解方法更简洁高效，是LLM压缩与加速的新范式
arXiv:2606.08565v1 Announce Type: new Abstract: Tensor networks provide efficient representations for compressing large neural networks. By carefully …
揭秘分层分解+监管控制，告别大模型上下文窗口爆炸，高效编排自主子代理。
We have all hit the "monolithic LLM wall." You design an incredibly capable AI agent, arm it with a suite of tools, and give it a complex, multi-step …
提出动态logit级门控机制，优化LLM专家路由效率，大幅降低推理成本。
arXiv:2606.04378v1 Announce Type: new Abstract: Leveraging multiple specialized LLMs can combine complementary strengths, but existing approaches trad…
不用重新训练就能替换大语言模型，MIT的MeMo方案让性能飙升26%，打破向量数据库局限
Enabling LLMs to acquire new knowledge after training remains a major hurdle for enterprise AI — current solutions are either too expensive, too slow,…
提出Diet-LLM方法，通过融合任务特定重要性分数实现维度级全局剪枝，在保持性能的同时显著压缩大模型规模。
arXiv:2603.23985v3 Announce Type: replace Abstract: Large language models (LLMs) have demonstrated remarkable capabilities, but their massive scale po…
AI产品工程师求职：精通自主代理工作流、RAG管道与低延迟推理，曾构建多模态智能体Jarvis。
About me: I am an AI Product Engineer specializing in building autonomous agentic workflows. Recently, I built 'Jarvis', a multimodal autonomous agent…
探索统一自蒸馏框架UniSD，为大型语言模型的高效优化与性能提升提供新思路。
arXiv:2605.06597v2 Announce Type: replace Abstract: Self-distillation (SD) offers a promising path for adapting large language models (LLMs) without r…
提出数字熵损失函数，专门提升大语言模型对数字的处理能力，让AI更懂数值。
arXiv:2605.20369v1 Announce Type: cross Abstract: Number prediction stands as a fundamental capability of large language models (LLMs) in mathematical…
EpiCache提出场景式KV缓存管理，在资源受限设备上高效支持超长对话，内存优化有新招。
arXiv:2509.17396v4 Announce Type: replace Abstract: Modern large language models (LLMs) extend context lengths to millions of tokens, enabling coheren…
统一API实现文本参数优化，单LLM系统跨领域媲美专业工具，无需定制即可高效搜索
arXiv:2605.19633v1 Announce Type: cross Abstract: Can a single LLM-based optimization system match specialized tools across fundamentally different do…