Codex 核验（2026-10-09） 本篇为原日期采集快照。Stars、points、评论数、排序和截断摘要均不是截至核验日的实时统计；GitHub“作者”是仓库拥有者，HN“作者”通常是投稿账号，arXiv 字段可能只列部分作者，应以链接原文署名为准。 Mistral 官方 2026-10-06 公告称 Large 4 现处于 API 公测，权重计划当月底发布；截至 2026-10-09 不能据“开放权重”表述认定权重已可下载或自部署。原来源末尾多余反斜杠已移除。

2026-10-06 互联网学习扫描

扫描日期: 2026-10-06 总条数: 26 条来源: V2EX + HackerNews + arXiv + GitHub Trending

thedotmack / claude-mem

来源: https://github.com/thedotmack/claude-mem
来源类型: github-trending-daily
作者: thedotmack
热度: 536
评分: 9
主题: ai-agent
摘要: Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More

cathrynlavery / diagram-design

来源: https://github.com/cathrynlavery/diagram-design
来源类型: github-trending-daily
作者: cathrynlavery
热度: 227
评分: 9
主题: ai-agent, tools
摘要: Editorial diagram design for Claude Code, Codex, GitHub Copilot, Factory Droid, and Pi. 42 diagram types. Self-contained HTML + SVG. No shadows. No Mermaid slop.

CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling

来源: https://arxiv.org/abs/2610.06829v1
来源类型: arxiv-cs-ai
作者: Zeyuan Chen
评分: 8
主题: ai-agent, tools, ops
摘要: Open-source web agents are now strong enough to execute realistic browser tasks, but training them with reinforcement learning still depends on weak supervision: binary task success is too sparse for credit assignment, while frontier-language-model judges are too expensive to call at every step and

TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts

来源: https://arxiv.org/abs/2610.06824v1
来源类型: arxiv-cs-ai
作者: Dane Sherburn
评分: 8
主题: ai-agent, ops, papers
摘要: We introduce TasteVal, a benchmark to evaluate the experimental research taste of frontier models. We define research taste as the ability to pick interesting problems to solve, design experiments, and interpret experimental results. TasteVal measures the experimental component of research taste; gi

Base Models Can Reason By Taking a Cue From Training Data

来源: https://arxiv.org/abs/2610.06851v1
来源类型: arxiv-cs-ai
作者: Alexei A. Efros
评分: 7
主题: ai-agent, ops, papers
摘要: In this paper, we study how training data creates associations between the tokens at the start of a base model’s response and the reasoning behavior that follows. First, we demonstrate that fixing particular starting token cues makes a base model’s performance competitive with that of its reinforcem

MemPilot: Orchestrating On-Demand Multimodal Memory Curation for LLM Agents

来源: https://arxiv.org/abs/2610.06830v1
来源类型: arxiv-cs-ai
作者: Wenya Wang
评分: 7
主题: ai-agent, tools, ops
摘要: Memory has become integral to the LLM agent ecosystem, supporting information retention and reuse across interactions. However, most existing agent memory systems construct memory in a query-agnostic manner, which can incur unnecessary preprocessing cost and discard details that later prove essentia

Back to the Future: Rethinking EDA Infrastructure for Agentic Systems in Chip Design Verification

来源: https://arxiv.org/abs/2610.06790v1
来源类型: arxiv-cs-ai
作者: Thomas Karpati
评分: 7
主题: ai-agent, tools, ops
摘要: The unprecedented computational scale of modern artificial intelligence depends on complex, multi-billion-transistor Systems-on-Chip, yet the workflows that verify these chips remain stubbornly manual. Although Large Language Models (LLMs) have made rapid inroads into Electronic Design Automation (E

mattpocock / skills

来源: https://github.com/mattpocock/skills
来源类型: github-trending-daily
作者: mattpocock
热度: 972
评分: 7
主题: ai-agent
摘要: Skills for Real Engineers. Straight from my .agents directory.

morluto / rea

来源: https://github.com/morluto/rea
来源类型: github-trending-daily
作者: morluto
热度: 2963
评分: 7
主题: ai-agent
摘要: Reverse engineer anything with agents, from app behavior down to native binaries.

deepseek-ai / DeepGEMM

来源: https://github.com/deepseek-ai/DeepGEMM
来源类型: github-trending-daily
作者: deepseek-ai
热度: 363
评分: 7
主题: ai-agent, tools, ops
摘要: DeepGEMM: clean and efficient BLAS kernel library on GPU

msitarzewski / agency-agents

来源: https://github.com/msitarzewski/agency-agents
来源类型: github-trending-daily
作者: msitarzewski
热度: 621
评分: 7
主题: ai-agent
摘要: A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.

Tapo (Rust/Python library) now speaks TP-Link’s TPAP protocol

来源: https://mihai.dinculescu.dev/posts/tapo-speaks-tpap/
来源类型: hn-frontpage
作者: faithraven
热度: 121
评分: 6
主题: tools

One Figure, Every Canvas: Editable Flowchart Relayout via Agentic Pipeline

来源: https://arxiv.org/abs/2610.06852v1
来源类型: arxiv-cs-ai
作者: Yu-Lun Liu
评分: 6
主题: ai-agent, tools, papers
摘要: Pipeline figures in ML papers must be repurposed across many canvases, including paper columns, 16:9 slides, portrait posters, 1:1 social teasers, 9:16 phone previews. Each format imposes a different aspect ratio on the same computational graph, where any silently broken connection misrepresents the

Learning to Read the Contextual Tokens in Diffusion Transformers

来源: https://arxiv.org/abs/2610.06844v1
来源类型: arxiv-cs-ai
作者: Or Patashnik
评分: 6
主题: ai-agent, tools
摘要: Multimodal Diffusion Transformers (MM-DiTs) jointly process visual and textual representations throughout generation. These models repeatedly update the text tokens through multimodal attention, forming dynamic contextual tokens whose function is not well understood. In this work, we introduce a fra

Recursive Video In-Context Learning for Agentic Robot

来源: https://arxiv.org/abs/2610.06843v1
来源类型: arxiv-cs-ai
作者: Yuzhang Shang
评分: 6
主题: ai-agent, tools
摘要: LLM agents that orchestrate frozen vision-language-action (VLA) policies improve across episodes through text memory, which records what the agent did but not how the task is done. A demonstration video shows it, but fits poorly into an agent’s context. The full video slows every turn, fixed keyfram

MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers

来源: https://arxiv.org/abs/2610.06801v1
来源类型: arxiv-cs-ai
作者: Chunchao Guo
评分: 6
主题: ai-agent, tools, papers
摘要: Sparse attention is a primary approach to reducing the latency of diffusion transformers in long-sequence generation tasks, such as video and high-resolution 3D asset generation. However, existing methods can degrade generation quality and fidelity at high sparsity levels. Through controlled oracle

tester-army / e2e

来源: https://github.com/tester-army/e2e
来源类型: github-trending-daily
作者: tester-army
热度: 1720
评分: 6
主题: tools
摘要: Next generation e2e testing framework for web and mobile apps.

boykopovar / AnyPS5

来源: https://github.com/boykopovar/AnyPS5
来源类型: github-trending-daily
作者: boykopovar
热度: 943
评分: 6
主题: ai-agent, ops
摘要: Tool for automatic PS5 executables porting to Linux and Windows

DuarteSantos8 / openGym

来源: https://github.com/DuarteSantos8/openGym
来源类型: github-trending-daily
作者: DuarteSantos8
热度: 1419
评分: 6
主题: ai-agent
摘要: Self-hosted gym & body-weight tracker — plan routines, log workouts (supersets, warm-ups, cardio), see which muscles are trained, fatigued or detrained, import from FitNotes/Strong/Hevy, passkey login. Your data, your server.

开源了 12 个电商短视频 Agent Skills：一张商品图 → 带配音字幕的 15 秒竖屏视频，本地 ffmpeg 出片

来源: https://www.v2ex.com/t/1246694
来源类型: v2ex
作者: boisy
评分: 5
主题: ai-agent, tools
摘要: 之前发过一个电商出图的 Skills ，有人问能不能接着把图变成视频。这周把这部分也做完了，开源（ MIT ）： https://github.com/xianyu110/ecommerce-video-skills?utm_source=v2ex&utm_medium=post&utm_campaign=ecommerce-video-skills&utm_content=body 想解决的问题：让大模型写带货分镜很容易，但写完还得自己开剪映一镜一镜拖。我想让 agent 不光写脚本，还能把片子真剪出来。做法：让 agent 输出一份 storyboard.json （每镜：用哪张图、

Mistral Large 4

来源: https://mistral.ai/news/mistral-large-4/
来源类型: hn-frontpage
作者: Philpax
热度: 1413
评分: 5
主题: ai-agent
摘要: https://docs.mistral.ai/models/mistral-large-4-0

Dust: Pretraining Transformers Without Backpropagation

来源: https://qlabs.sh/research/dust
来源类型: hn-frontpage
作者: E-Reverance
热度: 263
评分: 5
主题: ai-agent

earthtojake / text-to-cad

来源: https://github.com/earthtojake/text-to-cad
来源类型: github-trending-daily
作者: earthtojake
热度: 620
评分: 5
主题: ai-agent
摘要: Give your agent CAD superpowers.

pbakaus / impeccable

来源: https://github.com/pbakaus/impeccable
来源类型: github-trending-daily
作者: pbakaus
热度: 609
评分: 5
主题: ai-agent
摘要: The design language that makes your AI harness better at design.

ayghri / i-have-adhd

来源: https://github.com/ayghri/i-have-adhd
来源类型: github-trending-daily
作者: ayghri
热度: 318
评分: 5
主题: ai-agent
摘要: A skill to stop your coding agent from burying the answer. ADHD-friendly output.

知识库

探索

2026-10-06 互联网学习扫描

2026-10-06 互联网学习扫描

thedotmack / claude-mem

cathrynlavery / diagram-design

CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling

TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts

Base Models Can Reason By Taking a Cue From Training Data

MemPilot: Orchestrating On-Demand Multimodal Memory Curation for LLM Agents

Back to the Future: Rethinking EDA Infrastructure for Agentic Systems in Chip Design Verification

mattpocock / skills

morluto / rea

deepseek-ai / DeepGEMM

msitarzewski / agency-agents

推荐一个自带 TEE 防伪证明的 API 聚合中转站 wokey.ai

Tapo (Rust/Python library) now speaks TP-Link’s TPAP protocol

One Figure, Every Canvas: Editable Flowchart Relayout via Agentic Pipeline

Learning to Read the Contextual Tokens in Diffusion Transformers

Recursive Video In-Context Learning for Agentic Robot

MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers

tester-army / e2e

boykopovar / AnyPS5

DuarteSantos8 / openGym

开源了 12 个电商短视频 Agent Skills：一张商品图 → 带配音字幕的 15 秒竖屏视频，本地 ffmpeg 出片

Mistral Large 4

Dust: Pretraining Transformers Without Backpropagation

earthtojake / text-to-cad

pbakaus / impeccable

ayghri / i-have-adhd

关系图谱

目录

设置