GitHub Repo

GitHub 熱門專案、開源工具、框架、runtime 與重要 release。

GCache Allocates Diffusion Caching Based on End-to-End Error Impact, Cutting Wan2.1 LPIPS to 0.0316 at the Same Speed

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.

2026-08-14

推論系統

vToken Reclaims KV Cache Through Token-Level Virtualization, Boosting Throughput by Up to 37% on a Single H100

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.

2026-08-14

AI 基礎設施與訓練系統

LazyTrain Overlaps Memory Transfers With Mixed-Integer Scheduling, Speeding Up 27B Model Training on a Single H800 by 1.24×

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.

2026-08-14

推論系統

Trie Automata Precompute Token Masks, Delivering Up to 29× XGrammar’s Throughput for Finite-Choice Decoding in vLLM

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.

2026-08-14

最新模型

混合線性注意力在完整注意力層前形成巨量啟用尖峰，規律橫跨 1.2B 至 397B 模型

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。

2026-08-14

代理工具

Model Main Effects Fall Below 3% Across Three Agent Benchmarks, as DDR Recasts Leaderboards as a Deployment Reliability Problem

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.

2026-08-14

代理工具

LLM Agents Send HPC Workloads to Heterogeneous Clusters, with Descriptive Hardware Data Raising the Success Rate from 48% to 87%

An LLNL team enabled agents to explore resources managed by Flux, Slurm, Kubernetes, and other systems, then translate high-level intent into verifiable job specifications. Experiments across six cloud clusters show that architecture, network, and memory metadata mainly rule out machines incapable of running a workload, rather than serving as a reliable performance optimizer.

2026-08-13

代理工具

GSR Compiles Scoring Rules into Typed Graphs, Improving Exact-Score Agreement for LLM Judges by Up to 6.75 Points

Graph-Structured Rubrics compile natural-language rubrics into typed judgment, transformation, aggregation, and gating nodes before seeing the response to be evaluated. Experiments with GPT‑OSS‑120B show that GSR outperforms flat Prometheus-style scoring, although no implementation has been released and the gains vary substantially by dataset.

2026-08-13

最新模型

64-Way Self-Consistency Voting Actually Reduces Per-Question Accuracy on Most Hard GPQA Problems

A preregistered study found that, on most GPQA Diamond questions, Qwen2.5-7B and Llama-3-8B become increasingly likely to lock onto the wrong answer as more samples are put to a vote. Overall average scores still rise slightly, showing how aggregate metrics can conceal per-question regressions.

2026-08-13

最新模型

MiniMax H3 Releases 33B Base Weights with Native Synchronized Video and Stereo Audio Generation

MiniMax H3 uses a single Transformer to jointly predict video and audio latents, supporting generation from text, first and last frames, and multimodal references. The open release can generate 768p audiovisual content locally, but the official 2K workflow still relies on two proprietary cloud modules.

2026-08-13

最新模型

LTX‑2.5 Updates Its Open Video Model With a Diffusion Decoder and Single-Pass Multi-Shot Generation

LTX‑2.5 releases trainable 22B weights and eight-step distilled weights, adding character, audio, and scene consistency across shots. The complete quick-start components total roughly 66 GiB, Diffusers support still requires installing from the main branch, and the model is subject to a community license with a revenue threshold.

2026-08-13

最新模型

Qwen3.8 Makes 2.4T-Parameter Max-Class Weights Available for the First Time, Activating 95B per Token

Alibaba’s Qwen team has released Qwen3.8-2.4T-A95B, featuring 512 experts and a hybrid of linear and standard attention, with support for contexts of up to approximately 1.01 million tokens. The model offers adjustable reasoning depth, but accepts text only and cannot disable thinking mode, making its practical deployment requirements far higher than those of typical open-weight models.

2026-08-13