微信内可能无法直接打开本站。请点右上角 ··· → 在浏览器打开,或复制链接后用系统浏览器访问。
XMT
短闻
信流 · 上滑连读 · 来源可核
1 / 24
马斯克带火的太空算力领域,传一个 800MB 模型上天要四周
「所有劝我不要干太空算力的朋友,现在全在干太空算力。」 太空字节联合创始人秦贵军这样描述过去一年。这家公司 2024 年底成立,做的是太空数据中心;他本人参与过北斗导航重大专项,在卫星行业干了十几年。…
- 那些劝过他的朋友也大多是同一个画像:传统航天出身,两年前还觉得把算力送上天是个不着调的故事,如今已经用脚投票。
- 他画过一张国内太空算力的行业热力图,从长三角到北京亦庄、中关村,再到成都、西安、合肥,玩家已经超过三十家,正在往四十家走。
- 2026 年 6 月 12 日,SpaceX 以每股 135 美元、约 1.
RSS 官方收录 · 可信分层展示
A decodability criterion predicts when hidden-state selection beats majority voting in large language models
arXiv:2608.…
- 17124v1 Announce Type: new Abstract: Combining the answers a large lan…
- Voting is unreliable on difficult questions, where the sampled answers…
- Selecting a candidate by reading a correctness signal from the model's…
RSS 官方收录 · 可信分层展示
DiSCO: Defending text-to-image generation through distribution-guided contrastive prompt optimization
arXiv:2608.…
- 17067v1 Announce Type: new Abstract: As text-to-image generative model…
- Existing defenses predominantly operate under white-box assumptions, r…
- Black-box alternatives based on LLM prompt rewriting offer broader app…
RSS 官方收录 · 可信分层展示
KernelArc: A Multi-Agent Framework for GPU Kernel Optimization
arXiv:2608.17071v1 Announce Type: new Abstract: We present KernelArc, a multi-agent framework for autonomous GPU kernel optimization across heterogeneous workloads.…
- Strategy-specialized agents run in parallel and coordinate through con…
- We evaluate \kernelarc{} on NVIDIA H100 and B200 GPUs using category-r…
- The resulting implementations span custom BF16 GEMM, static cuBLASLt E…
RSS 官方收录 · 可信分层展示
Memory Is Communication: The Frontier Between Remembering and Signaling
arXiv:2608.17053v1 Announce Type: new Abstract: A bounded agent may obtain information for a decision from its own past, from peers, or from both sources.…
- Retaining task-relevant history can reduce later communication, while …
- Under limits on both resources, how should an agent allocate its infor…
- Given a fixed task and decision rule, the memory and message rate pair…
RSS 官方收录 · 可信分层展示
SkillEffect: Checked Lowering for Memory-Bounded Agent Tools
arXiv:2608.…
- 17007v1 Announce Type: new Abstract: Agent Skills can specify procedur…
- However, when models turn this guidance into code for existing tool in…
- We present SkillEffect, a checked-lowering runtime for computations wi…
RSS 官方收录 · 可信分层展示