2026年9月15日 技术热点总结

项目简介：Free open-source extractor for AI coding assistant chat histories. Supports Claude Code, Cursor, Windsurf, Aider, Cline/Roo Code, and more.

技术栈：Python、ai、ai-data-extraction、claude-code、cursor

核心介绍：Extract your own local chat history from AI coding assistants into a single, normalized JSONL format – for fine-tuning, personal analytics, or just backing up years of conversations before an app’s local database gets cleared. Auto-discovers and extracts complete conversation history, including: tool’s storage actually contains Every script searches macOS, Linux, and Windows…

项目简介：The language app you eventually delete. A native iPhone companion for learning through conversation.

技术栈：Kotlin、ios、language-learning、open-source、swiftui

核心介绍：Mural is a native iPhone and Android app for learning through conversation. Speak to a warm, animated orb, follow the meaning when you need it, and practise words again in later conversations. Mural adjusts the challenge from the evidence in your replies.

项目数据：⭐ 783 Stars，🍴 169 Forks

项目简介：Official Project Page for Recurrent Looped Transformer (RLT)

技术栈：HTML

核心介绍：[Paper] [Project Website] 1. Latent reasoning with infinite temporal depth. Each processed token extends a recurrent path through the decoder. With decoder depth $L_D$, the path traverses $tL_D$ blocks after $t$ tokens, while the number of blocks executed per token stays fixed.

项目数据：⭐ 743 Stars，🍴 77 Forks

项目简介：Learning notes and tooling skills for AI video – AI 视频相关的学习与工具 skill

技术栈：HTML、ai-video、claude-skills、ffmpeg、shot-analysis、video-analysis

核心介绍：视频侧的 Claude Code / Codex skill。 我建了一个付费AI视频交流群，聊 AI 视频的工作流、工具和实操。 有兴趣的加我：微信 hao_dev，添加时备注 github。

项目数据：⭐ 666 Stars，🍴 89 Forks

技术栈：Python

核心介绍：proxy instances (Lunel Core) from a single web console (Lunel Console) — without ever touching a server. Derived from the RVG Gateway relay engine, refactored into a clean product architecture: real deployments, real health checks, real isolation, real endpoints. No simulated states, no fake buttons. GitHub → Lunel Console → Create Instance → Deploy → Running → Endpoint → Manage link can never attach to another link’s session stream. per-link session caps.

项目数据：⭐ 662 Stars，🍴 1,580 Forks

项目简介：A full retained fly-connectome simulation with experimental memory and guarded Coinbase AgentKit trading actions.

技术栈：Python

核心介绍：A fly-connectome simulation that can operate a crypto trading account. Actual neural output, actual Coinbase integration. Profitable learning has not been demonstrated.

项目数据：⭐ 646 Stars，🍴 105 Forks

Reacting to sudden physical hazards (catching a slipping plate, dodging a falling knife) is both a meaningful test of embodied intelligence and a hard requirement for deploying multimodal large language models (MLLMs) as the decision coreof household robots. Existing evaluations, however, probe intuitive physics passively through question answering over videos, or target deliberate, long-horizon tasks such as navigation and rearrangement; none measure whether a model can turn physical understanding into immediate, safety-critical action. We introduce ReactHuman, the first physics-grounded b…

In hybrid forecasting, a language model is often one of several available signals. A system may already have a market, crowd, or statistical forecast and must decide whether the model adds useful information or should be ignored. The relevant target is therefore not standalone model accuracy, but relative competence, defined as the model's marginal value beyond the available external forecast. Under Brier loss, we characterize when model disagreement can improve an external forecast and derive the gain from using domain-specific rather than global pooling weights. We then introduce a compet…

Objects in post-fire environments often undergo irreversible physical transformations that change their geometry, material state, and visual appearance. Detecting and identifying these remnants is critical for locating hazards, reconstructing pre-incident contents, and inventorying losses. Unlike standard image corruptions, these degradations affect the physical structure of the object itself. To study this setting, we introduce TRACE, a transformation-aware benchmark for post-fire object understanding. TRACE contains 21.4K real-image-grounded synthetic scenes and paired object-level pristi…

Large language models are unreliable at arithmetic, which is a problem for clinical calculators where a single numerical error changes the recommendation. The standard response is to hardcode each calculator as a validated function, one at a time. We test an alternative: the model does not calculate. Instead, it writes case-specific Python that a restricted local executor runs as a deterministic solver, and the model's task reduces to deciding how to use it. We evaluate this Program-Solve interface on MedCalc-Bench Verified (1,100 cases, 55 calculators) against direct model arithmetic and a…

We present our submission to the EgoProactive track of the ECCV 2026 Wearable AI Challenge, which ranked first in the large-model division and second in the <=2B division. The task requires a wearable assistant to decide after each eight-second segment of egocentric video whether to intervene or remain silent.
Our approach has two main components. First, we reformulate intervention timing as single-token classification. Rather than generating either interrupt<utterance> or silent, the model predicts yes or no, and we derive the decision from the renormalised probabilities of these two tok…

We describe our entry to the EgoLongQA track of the Wearable-AI Challenge in ECCV 2026, which placed first in the <=2B parameter division with 0.8279 on the held-out test set. Our system is a single 2B vision-language model that answers multiple-choice questions about ten-minute egocentric videos in one greedy forward pass; It is obtained by distilling the junior perception module of a tool-using agentic pipeline, not the agent itself into a small student, using teacher traces filtered to those that answered correctly. it reaches 89% of the accuracy of the large agentic pipeline using 1.1% …