2026年9月3日 技术热点总结

项目简介：Dress AI Sponsor

技术栈：Python、18comic、coomer、coomer-downloader、coomer-party、coomer-porn、coomer-su、doujin-moe、ehviewer

核心介绍：🚀 Turn your closet into an AI-powered fashion assistant. Upload your wardrobe, get smart outfit recommendations for any occasion, and visualize your looks with generative AI — all self-hosted, private, and free. 🎥 Demo Video · 📖 Documentation · 🐳 Quick Start · 💬 Discord Whether you’re a fashion enthusiast, a boutique owner, or a developer building the next generation of fashion tech, Dress AI Service gives you a complete, production-ready foundation…

项目简介：Turn any AI agent into a computational researcher. HexStellar Cortex delivers software-accelerated optimization, quantum computing, scientific computing, decision intelligence, and verifiable execution through a Python CLI and API—with certainty labels, verification receipts, examples, and a free sandbox. Start instantly: pip install hexstellar

技术栈：Python

核心介绍：> Distribution status： this repository is the public HexStellar CLI 1.0 source. The PyPI 1.x > distribution is activated in a separate, audited release step. Until > pypi.org/project/hexstellar reports a 1.x rel…

项目简介：DLSS 5 neural rendering in D3D11/D12/Vulkan games that ship without any DLSS — feeds a synthetic DLAA contract (ReShade depth + motion vectors) to the DLSS 5 add-on via a private D3D12 device.

技术栈：C++

核心介绍：DLSS 5’s neural-rendering add-on only works by hooking a game’s own DLSS calls. A game that has no DLSS never makes those calls, so the add-on sits idle. DLSS5-Feeder makes the calls itself. It builds a complete DLSS DLAA “contract” out of what ReShade already has — the frame being processed, the depth buffer, and estimated optical-flow motion vectors — runs a genui…

Dexterous manipulation policies learned by imitation are typically evaluated for robustness to variation in scenes, objects, or instructions, but their performance across task execution speeds is less often examined. This leaves open how much temporal robustness a learner retains relative to the expert it imitates. We compare an expert and learner under the same task conditions, initial-condition draws, and speedup factors. We instantiate the evaluation in ParcelStow, a contact-rich task in which the robot acquires, reorients, and inserts a parcel. The demonstrations span the speedup range …

Logit-based knowledge distillation (KD) is used to train smaller language models (LMs) via supervision from stronger teachers, but whether its benefits are consistent across training stages remains unclear. Through controlled experiments, we find that forward Kullback-Leibler (KL) distillation–the standard KD formulation–with post-trained teachers behaves fundamentally differently during mid-training, an intermediate phase of self-supervised learning on curated corpora. Surprisingly, while forward KD simultaneously improves reasoning and factual recall during pre-training relative to stan…

LLM judges are widely used to evaluate agentic tool-calling systems, yet their reliability on structured, dependency-driven workflows remains largely unexamined. We present AgentJudgeBench, the first benchmark to systematically study LLM-as-a-judge reliability for agentic tool-calling over workflow DAGs, as distinct from the broader LLM-as-a-judge task of open-ended text or preference evaluation. The benchmark comprises 3,808 instances spanning six DAG topologies and three difficulty tiers, evaluated with five generators (3B-70B open-weight models and GPT-5.4) and six judges (20B to frontie…

Valuable data remains embedded in unstructured sources: web pages, reports, contracts, filings, earnings calls, and PDFs. The big bet in enterprise AI is deploying LLM agents that reason over this data to answer complex questions for every knowledge worker. Agents can do this today, but at prohibitive cost. Each question repeatedly opens large documents to recover scattered evidence, consuming up to a million tokens. However, if the data were already structured, the same question would reduce to a cheap database lookup. For example, on FanOutQA benchmark, reasoning over an ideal pre-structu…

Test-time adaptation (TTA) typically assumes that model parameters can be updated at inference time. This assumption is restrictive for inference-only accelerators, frozen or third-party models, and memory-constrained deployments, and standard BatchNorm-based TTA configurations may also become inactive on architectures without BatchNorm. We study adaptation when the learned model must remain frozen. We introduce CASTER, a gradient-free method that stores source class statistics in a discriminative subspace, estimates a class-shared affine transformation from target-batch moments, and analyt…

Commercial short-drama production follows a multi-stage chain: script, storyboard, keyframe imagery, shot-level video, and the finished short drama. Most existing benchmarks evaluate solely the video-generation stage using pre-authored inputs instead of real upstream pipeline outputs. This leaves two critical questions unanswerable: whether each stage adheres to the original script intent (rather than only its immediate input prompt), and whether disparate shots remain coherent after assembly into multi-episode releases. We present DramaChain Bench, the first short-drama benchmark that eval…