1
R 星《GTA6》地图又泄露了，但这次与大葱哥无关
《GTA6》地图再度泄露，实机画面与细节曝光，黑客组织又搞事，吃瓜别错过！
IT之家 8 月 21 日消息，据游戏媒体 Insider Gaming 今天报道，R 星《GTA6》游戏再度遭遇泄露事件，本作的地图过去几个小时内在社交媒体广泛传播， 但这次事件与暂时陷入沉寂的“大葱哥”（Cyberleek）并没有直接关联 。 据报道，这批《侠盗猎车手 VI》泄露数据至少在网络上…
2
Kozuchi Agent: A Language-Agnostic Open-Weight Agent for Software Repair
开源权重、语言无关，Kozuchi Agent 为软件自动修复带来跨语言新范式。
arXiv:2608.15579v1 Announce Type: cross Abstract: Industrial software-engineering teams increasingly need LLM agents that turn bug reports into correc…
3
Understanding Calibration and Truncation Error Propagation in Training-Free Low-Rank Compression for LLMs
深入剖析免训练低秩压缩中校准与截断误差的传播机制，为LLM高效压缩提供理论支撑。
arXiv:2608.08506v1 Announce Type: new Abstract: Training-free low-rank compression frameworks have been gaining prominence for LLM compression given t…
4
Prune Once: Retraining-Free Task-Agnostic Pruning for Vision-Language Models
视觉语言模型如何轻量化？这项研究提出免重训练的任务无关剪枝方案，一次剪枝即可适配下游任务。
arXiv:2608.06901v1 Announce Type: cross Abstract: Vision-language models (VLMs) have achieved remarkable generalization across diverse multimodal task…
5
Show HN: Pacific Slate: a self-hosted, model-agnostic multi-agent AI assistant
自托管多代理AI助手，模型无关且能学习偏好，把私有数据变成你的专属智能管家
I didn't want to buy a standalone computer or repurpose a laptop to run constantly so I could maintain a system to sync my LLMs, so I built this. It's…
6
Minnow – open-source AI development Workspace
免费开源的AI开发工作空间，内置114个工具，可接入任意模型，灵活强大
Article URL: https://github.com/HenriGrimm/Minnow Comments URL: https://news.ycombinator.com/item?id=49171199 Points: 2 # Comments: 1
7
8
ChainMark: Model-Free LLM Watermarking with Closed-Form Calibration
无需访问模型内部，用闭式校准实现LLM文本水印，防篡改可溯源。
arXiv:2607.18445v1 Announce Type: cross Abstract: Regulatory regimes such as the EU AI Act mandate machine-readable marking of synthetic text, but exi…
9
VarRate: Training-Free Variable-Rate KV Cache Compression for Long-Context LLMs
无需训练即可实现可变率KV缓存压缩，显著降低长上下文LLM推理内存开销，附大量实验验证效果。
arXiv:2607.15498v1 Announce Type: cross Abstract: The key-value (KV) cache is the main memory bottleneck in long-context large language model (LLM) in…
10
Open Science Desktop: A Local-First Experimental Tool for AI Research
开源本地优先的AI研究工具体验，模型无关且跨平台，Tauri构建轻量高效。
The launch of Open Science Desktop marks a notable entry into the AI research tool space, positioning itself as a local-first, model-agnostic workbenc…
11
More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges
自博弈评判者打分的是“像不像”而非“对不对”，参考无关LLM裁判存在结构性缺陷。
arXiv:2607.05904v1 Announce Type: new Abstract: Training a language model against its own reference-free judgments (the premise of self-rewarding, sel…
12
HiFA4: Training-Free 4-bit FlashAttention on Ascend HIF4 NPUs for LLM Inference
华为昇腾NPU上实现免训练4-bit FlashAttention，大幅降低LLM推理内存与计算开销。
arXiv:2607.04302v1 Announce Type: cross Abstract: We present HiFA4, a post-training operator-level design that executes both QK^T and PV in FlashAtten…
13
Show HN: Open Science, open-source alternative to Claude Science
对抗AI研究封闭化，开源、本地优先、模型无关的科研工作台，MIT许可，完全可复现。
Claude Science just launched. It’s a sign AI research is moving toward centralized, closed systems. So we built the opposite: Open Science: a local fi…
14
More AI – Open-source model-agnostic AI desktop
开源、模型无关的AI桌面工具，让你在本地自由切换调用不同大模型，告别平台绑定。
Article URL: https://github.com/DougTrier/MoreAI Comments URL: https://news.ycombinator.com/item?id=48755267 Points: 1 # Comments: 1
15
ReFreeKV: Towards Threshold-Free KV Cache Compression
新论文提出ReFreeKV，无需预设阈值的KV缓存压缩，实现无损内存缩减，解决现有方法依赖人工设定预算的痛点。
arXiv:2502.16886v4 Announce Type: replace-cross Abstract: To reduce memory consumption during LLM inference, a handful of methods have been proposed f…
16
How Shopify built an AI stack that doesn't care which models survive
Shopify构建了一套模型无关的AI堆栈，通过统一调度引擎自动切换大模型，实现工作流零中断与降本增效。
Shopify built an LLM proxy that gives every engineer access to multiple AI providers — with automatic failover when any one of them goes down, changes…
17
Explainable Control Framework (XCF) based on Fuzzy Model-Agnostic Explanation and LLM Agent-Supported Interface
融合模糊模型无关解释与LLM智能体交互，打造透明可解释的控制框架，为复杂系统决策提供新思路。
arXiv:2606.25941v1 Announce Type: cross Abstract: Increasing demand for precise and reliable control in complex scenarios has led to the development o…
18
A Test-time Actor-Critic Approach to News Images Generation
用Actor-Critic范式在测试时优化新闻配图生成，模型无关且即插即用，AI图像生成又有新思路。
arXiv:2606.21304v1 Announce Type: new Abstract: This paper introduces the CERTH-ITI solution for the MediaEval NewsImages 2026 challenge, which focuse…
19
QG-MIL: A Gated Transformer Aggregator for Domain-Agnostic Multiple Instance Learning in Medical Imaging
提出门控Transformer聚合器，实现医学影像领域无关的多实例学习，突破传统MIL方法局限
arXiv:2606.20027v1 Announce Type: new Abstract: Attention-based Multiple Instance Learning aggregators in medical imaging are prone to attention conce…
20
AIMER: Calibration-Free Task-Agnostic MoE Expert Pruning
无需校准的任务无关MoE专家剪枝方法，显著降低大模型推理成本。
arXiv:2603.18492v3 Announce Type: replace Abstract: Mixture-of-Experts (MoE) language models increase parameter capacity without proportional per-toke…