2026年7月5日 技术热点总结

项目简介：Everything I know about running LLMs locally

技术栈：Shell

核心介绍：Have $2k burning a hole in your pocket and want some local, state-of-the-art machine intelligence? How about $40k? If Dario and Altman are giving you heartburn (they should be), read on to figure out how to run this new kind of computing locally.

项目数据：⭐ 746 Stars，🍴 33 Forks

项目简介：GPU worker client for the Talos network. Pairs with your Talos account, serves open-model inference jobs over a WebSocket, and reports uptime for payouts.

技术栈：Python、ai、distributed-computing、gpu、llm、ollama、python、websocket、worker

核心介绍：Share your GPU with the Talos network and earn. This is the downloadable client (intended to live in its own public GitHub repo, talos-worker). It pairs with your Talos account using a code, then serves open-model inference jobs from the network via your local Ollama, reporting uptime and earning a share of real usage revenue. The Talos web …

项目简介：Always-on token-efficiency skill for coding agents (Claude Code, Codex, Cursor, Windsurf, Cline). ~31% lower bill on average, no loss of correctness.

技术栈：Shell

核心介绍：Always-on token-efficiency skill for coding agents — Claude Code, Codex, Cursor, Windsurf, Cline. Trims tokens across the whole session (replies, docs, tests, code, context, tool use) without losing correctness. ≈31% lower bill on average (−17% to −54% by session type) and −30% to −81% output on real Sonnet 5 runs.

项目数据：⭐ 452 Stars，🍴 0 Forks

技术栈：Astro

核心介绍：English | 中文 A lightweight, SEO-friendly, bilingual (EN / 中文) single-page tool that scans your user. One click runs an animated scan of each signal, the gauge climbs as risk adds up, and you get a verdict plus the list of matched signals. Everything runs 100% locally — no network requests, no data upload. Built with Claude Fable 5. When Claude Code is pointed at a non-official endpoint via ANTHROPIC_BASE_URL, it

项目数据：⭐ 372 Stars，🍴 36 Forks

Grid-based approaches to approximate nearest neighbor (ANN) search have been absent from modern scaling analyses. We present a systematic characterization of a multiprobe grid algorithm with respect to dataset size N and dimensionality d. Our experiments reveal a previously unreported d-scaling crossover on the GloVe embedding family, in which multiprobe grid search maintains an approximately constant dimensional scaling exponent while other graph-, tree-, and partitioning-based methods exhibit degrading throughput. The advantage comes with near-linear query scaling in N, but also with lowe…

Traffic matrices (TMs) capture network-wide origin-destination demand and are central to traffic engineering, yet accurate whole-matrix forecasting remains challenging when prediction must be performed under the memory, update, and training-budget constraints of online network control. This paper investigates whether compact quantum-inspired recurrent models can provide effective TM forecasts without relying on dedicated graph, transformer, or diffusion modules. We adapt gated quantum-inspired Kolmogorov-Arnold network fast-weight programmers (QKAN-FWPs) to direct multi-step Abilene TM fore…

Foundation models are routinely released to the public, yet the data recipes used to train them — such as domain mixture weights that determine how different sources are sampled — are rarely disclosed. This creates an access asymmetry: researchers study the resulting models but lack visibility into the training distribution that produces them. Prior works for inferring training data, such as membership inference, detect at the level of individual samples and thus cannot characterize the global composition of the training corpus. We introduce WARP, a framework that recovers a fine-tuned mo…

Memory expertise is a learned skill: knowing what to encode, when to retrieve, and how to organize knowledge–a capacity known in cognitive science as metamemory. We bring this perspective to LLMs by treating memory management as a trainable skill. We promote file-system operations to first-class memory actions alongside task actions, letting the model itself decide how to manage its memory. This memory skill improves along two axes: the structure that supports it (prompts, file schemas, action vocabulary), and the proficiency of the model exercising it. Both axes resist manual optimization…

Large Language Model (LLM)-based agents can solve complex procedural tasks by interacting with environments over multiple turns, but this ability typically depends on large models, long contexts, and repeated inference calls. This makes advanced memory-augmented agents difficult to deploy on resource-constrained devices. We introduce DuoMem, a dual-space distillation framework that transfers procedural problem-solving ability from a large teacher model to compact student models. DuoMem distils in two complementary spaces: (1)context-space distillation, which replaces student-generated memor…

In long-context use, large language models frequently synthesize answers from the meaning of a relevant context span rather than literally copy-pasting them. Identifying which attention heads perform this synthesis matters for interpreting long-context model behavior. Yet existing detectors miss these heads by construction: they reward heads whose attended token matches the generated token, a literal-copy criterion that captures where a head reads but not what it writes through its output-value (OV) circuit, the very mechanism that carries non-literal retrieval. We introduce Logit-Contribut…