Archive
新模型、開放權重、多模態、語音、影像與模型架構。
推論系統
Rather than balancing only tokens or the number of active experts, TEMPO estimates both HBM weight-loading costs and GEMM tile padding. An SGLang prototype improves throughput by 4% to 6% on some Qwen3-235B workloads, but adds only overhead on the communication-bound DeepSeek-V3.
代理工具
OmniScientist lets agents directly inspect images, signals, audio, video, 3D structures, and trajectories before completing ideation, experimentation, and paper writing. All 36 cases produced complete manuscripts, but the case count, code availability, and automated evaluation design still require clarification.
代理工具
StateBridge uses an orthogonal Procrustes transformation to map the sending agent’s final-layer hidden states into the receiving agent’s input embedding space, without training a projector. Its average scores across four models exceeded the strongest baseline by 2.4 to 2.9 points, but it currently supports only agents using identical model weights.
推論系統
FlashDrive tackles four bottlenecks—visual encoding, LLM prefill, reasoning tokens, and trajectory denoising—cutting Alpamayo 1.5-10B single-window latency on an RTX PRO 6000 from 717 to 151 milliseconds. The open-source implementation approaches real-time inference, but requires specialized fine-tuning, and its closed-loop evaluation covers only 100 simulated clips.
最新模型
AlayaWorld has rewritten the spatial memory and conditioning pipeline of its interactive world model so that camera controls, historical frames, and generated content use aligned causal VAE representations. A new report claims the highest overall consistency score on WBench, but the updated weights and inference code have yet to be released.
推論系統
GCache no longer decides whether to recompute based on local similarity between adjacent denoising steps. Instead, it searches for a cache-reuse schedule with the lowest impact on final image quality. On Wan2.1, it substantially reduces deviation while maintaining a 2.17× speedup, but policy search requires multiple H100 GPUs and the implementation has not yet been released.
推論系統
vToken adds a token-to-physical-location mapping layer on top of PagedAttention blocks, compacting scattered KV gaps left after eviction. Its vLLM prototype reduces retained blocks by up to 72.3%, though testing remains limited to a single machine with one GPU and primarily 7B- and 8B-class models.
AI 基礎設施與訓練系統
Before training, LazyTrain jointly determines where activations should be saved, offloaded, or recomputed, fitting PCIe and NVMe transfers into GPU compute windows. In Qwen3.6-27B experiments, performance increased from 176.90 to 219.95 TFLOPS, although the system has so far been validated only on a single GPU with offline scheduling.
推論系統
A new method precomputes trie states and token masks for thousands to tens of thousands of valid strings, avoiding the cardinality bottleneck of general-purpose grammar compilation. The authors measured 219 requests per second at a batch size of 256, but have not yet released the implementation, and the figure also includes integration gains from bypassing the existing guided-decoding path.
最新模型
New research suggests that longer training windows may cause models to rely on context rather than encode knowledge in their parameters. Both pretraining experiments with models up to 750M parameters and Qwen3 fine-tuning experiments show that, as supporting information increases, robustness in no-context and conflicting-evidence settings declines.
新研究發現，混合線性注意力模型的巨量啟用並非隨機散布，而會在完整注意力層前形成尖峰，並於中間的線性層維持平台。團隊公開分析程式與受控模型，但完整論文圖表所用的逐筆輸入尚未全部釋出。
代理工具
A new study uses generalizability theory to decompose results from TheAgentCompany, τ²-bench, and AppWorld, finding that agent–task interactions far outweigh average differences between agents. The authors propose the DDR reporting framework, which requires model-selection decisions to disclose task difficulty, holdout validation, cost, and required sample size.