AI Market Intelligence Report — May 13, 2026

Auto-generated AI Tools & Agents market intelligence for Facorreia vault. Sources: arXiv (cs.CL, cs.AI, cs.LG), GitHub Trending, Hugging Face, CoinDesk AI.

🔬 arXiv Papers — cs.CL, cs.AI, cs.LG (Top 10)

1. AlphaGRPO: Unlocking Self-Reflective Multimodal Generation in UMMs via Decompositional Verifiable Reward

Authors: Runhui Huang, Jie Wu, Rui Yang, Zhe Liu, Hengshuang Zhao
Published: 2026-05-12T17:59:47Z
Categories: cs.CV, cs.AI, cs.LG
Link: https://arxiv.org/abs/2605.12495v1
Summary: In this paper, we propose AlphaGRPO, a novel framework that applies Group Relative Policy Optimization (GRPO) to AR-Diffusion Unified Multimodal Models (UMMs) to enhance multimodal generation capabilities without an additional cold-start stage. Our approach unlocks the model's intrinsic potential to...

2. LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues

Authors: Di Wu et al. (7 authors)
Published: 2026-05-12T17:59:34Z
Categories: cs.CL
Link: https://arxiv.org/abs/2605.12493v1
Summary: Long-term memory is crucial for agents in specialized web environments, where success depends on recalling interface affordances, state dynamics, workflows, and recurring failure modes. However, existing memory benchmarks for agents mostly focus on user histories, short traces, or downstream task su...

3. Pion: A Spectrum-Preserving Optimizer via Orthogonal Equivalence Transformation

Authors: Kexuan Shi et al. (6 authors)
Published: 2026-05-12T17:59:34Z
Categories: cs.LG, stat.ML
Link: https://arxiv.org/abs/2605.12492v1
Summary: We introduce Pion, a spectrum-preserving optimizer for large language model (LLM) training based on orthogonal equivalence transformation. Unlike additive optimizers such as Adam and Muon, Pion updates each weight matrix through left and right orthogonal transformations, preserving its singular valu...

4. Elastic Attention Cores for Scalable Vision Transformers

Authors: Alan Z. Song et al. (11 authors)
Published: 2026-05-12T17:59:26Z
Categories: cs.CV, cs.LG
Link: https://arxiv.org/abs/2605.12491v1
Summary: Vision Transformers (ViTs) achieve strong data-driven scaling by leveraging all-to-all self-attention. However, this flexibility incurs a computational cost that scales quadratically with image resolution, limiting ViTs in high-resolution domains. Underlying this approach is the assumption that pair...

5. Task-Adaptive Embedding Refinement via Test-time LLM Guidance

Authors: Ariel Gera, Shir Ashury-Tahan, Gal Bloch, Ohad Eytan, Assaf Toledo
Published: 2026-05-12T17:58:27Z
Categories: cs.CL, cs.IR, cs.LG
Link: https://arxiv.org/abs/2605.12487v1
Summary: We explore the effectiveness of an LLM-guided query refinement paradigm for extending the usability of embedding models to challenging zero-shot search and classification tasks. Our approach refines the embedding representation of a user query using feedback from a generative LLM on a small set of d...

6. Learning, Fast and Slow: Towards LLMs That Adapt Continually

Authors: Rishabh Tiwari et al. (9 authors)
Published: 2026-05-12T17:58:20Z
Categories: cs.LG, cs.AI
Link: https://arxiv.org/abs/2605.12484v1
Summary: Large language models (LLMs) are trained for downstream tasks by updating their parameters (e.g., via RL). However, updating parameters forces them to absorb task-specific information, which can result in catastrophic forgetting and loss of plasticity. In contrast, in-context learning with fixed LLM...

7. Beyond GRPO and On-Policy Distillation: An Empirical Sparse-to-Dense Reward Principle for Language-Model Post-Training

Authors: Yuanda Xu et al. (6 authors)
Published: 2026-05-12T17:57:48Z
Categories: cs.LG, cs.AI
Link: https://arxiv.org/abs/2605.12483v1
Summary: In settings where labeled verifiable training data is the binding constraint, each checked example should be allocated carefully. The standard practice is to use this data directly on the model that will be deployed, for example by running GRPO on the deployment student. We argue that this is often ...

8. ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents

Authors: Xuhao Hu et al. (9 authors)
Published: 2026-05-12T17:57:04Z
Categories: cs.AI
Link: https://arxiv.org/abs/2605.12481v1
Summary: Computer Use Agents (CUAs) can act through both atomic GUI actions, such as click and type, and high-level tool calls, such as API-based file operations, but this hybrid action space often leaves them uncertain about when to continue with GUI actions or switch to tools, leading to suboptimal executi...

9. OmniNFT: Modality-wise Omni Diffusion Reinforcement for Joint Audio-Video Generation

Authors: Guohui Zhang et al. (12 authors)
Published: 2026-05-12T17:56:59Z
Categories: cs.CV, cs.AI
Link: https://arxiv.org/abs/2605.12480v1
Summary: Recent advances in joint audio-video generation have been remarkable, yet real-world applications demand strong per-modality fidelity, cross-modal alignment, and fine-grained synchronization. Reinforcement Learning (RL) offers a promising paradigm, but its extension to multi-objective and multi-moda...

10. MEME: Multi-entity & Evolving Memory Evaluation

Authors: Seokwon Jung, Alexander Rubinstein, Arnas Uselis, Sangdoo Yun, Seong Joon Oh
Published: 2026-05-12T17:55:10Z
Categories: cs.LG, cs.CL
Link: https://arxiv.org/abs/2605.12477v1
Summary: LLM-based agents increasingly operate in persistent environments where they must store, update, and reason over information across many sessions. While prior benchmarks evaluate only single-entity updates, MEME defines six tasks spanning the full space defined by the multi-entity and evolving axes, ...

🚀 GitHub Trending AI/Agent Repos (Weekly)

| # | Repository | Description | Stars (wk) | |---|-----------|-------------|-----------| | 1 | anthropics/financial-services | Anthropic's financial services reference architecture | +13,176 | | 2 | Hmbown/DeepSeek-TUI | Coding agent for DeepSeek models in terminal (Rust) | +20,835 | | 3 | bytedance/UI-TARS-desktop | Open-Source Multimodal AI Agent Stack | +3,872 | | 4 | CloakHQ/CloakBrowser | Stealth Chromium for bot detection, Playwright replacement | +5,488 | | 5 | decolua/9router | Free AI coding proxy: 40+ providers for Claude/GPT/Gemini | +5,204 | | 6 | LearningCircuit/local-deep-research | Local deep research agent, ~95% on SimpleQA | +2,412 | | 7 | ruvnet/ruflo | Agent orchestration platform for Claude, multi-agent swarms | +7,088 | | 8 | VectifyAI/PageIndex | Vectorless, Reasoning-based RAG Document Index | +4,351 | | 9 | addyosmani/agent-skills | Production-grade engineering skills for AI coding agents | trending | | 10 | docusealco/docuseal | Open source DocuSign alternative | trending | | 11 | Imbad0202/academic-research-skills | Academic Research Skills for Claude Code | trending | | 12 | rohitg00/agentmemory | Persistent memory for AI coding agents | trending | | 13 | HKUDS/AI-Trader | 100% Fully-Automated Agent-Native Trading | trending | | 14 | TauricResearch/TradingAgents | Multi-Agents LLM Financial Trading Framework | trending | | 15 | InsForge/InsForge | Open-source backend platform for agentic coding | trending |

🤗 Hugging Face Recent Model Releases

LLM-OS-Models/gemma-4-E4B-Terminal-SFT-Native-Liquid-2Epoch | Pipeline: text-generation | Likes: 1 | Tags: transformers, safetensors, gemma4_text, text-generation
LLM-OS-Models/gemma-4-E4B-Terminal-SFT-Native-Liquid-1Epoch | Pipeline: text-generation | Likes: 1 | Tags: transformers, safetensors, gemma4_text, text-generation
LLM-OS-Models/gemma-4-E4B-it-Terminal-SFT-Native-Liquid-2Epoch | Pipeline: text-generation | Likes: 1 | Tags: transformers, safetensors, gemma4_text, text-generation
LLM-OS-Models/gemma-4-E4B-it-Terminal-SFT-Native-Liquid-1Epoch | Pipeline: text-generation | Likes: 3 | Tags: transformers, safetensors, gemma4_text, text-generation
LLM-OS-Models/gemma-4-E2B-Terminal-SFT-Native-Liquid-2Epoch | Pipeline: text-generation | Likes: 1 | Tags: transformers, safetensors, gemma4_text, text-generation
LLM-OS-Models/gemma-4-E2B-Terminal-SFT-Native-Liquid-1Epoch | Pipeline: text-generation | Likes: 1 | Tags: transformers, safetensors, gemma4_text, text-generation
LLM-OS-Models/gemma-4-E2B-it-Terminal-SFT-Native-Liquid-2Epoch | Pipeline: text-generation | Likes: 3 | Tags: transformers, safetensors, gemma4_text, text-generation
LLM-OS-Models/gemma-4-E2B-it-Terminal-SFT-Native-Liquid-1Epoch | Pipeline: text-generation | Likes: 1 | Tags: transformers, safetensors, gemma4_text, text-generation
mradermacher/Nayari-i1-GGUF | Pipeline: unknown | Likes: 0 | Tags: transformers, gguf, en, base_model:Crossie/Nayari
Ba2han/experimental2 | Pipeline: text-generation | Likes: 1 | Tags: transformers, safetensors, qwen3, text-generation

📰 AI News Highlights

Bitcoin miner MARA sold $1.5B of bitcoin to shift toward AI infrastructure — signaling the convergence of crypto mining infrastructure with AI compute demand (CoinDesk, May 12)
Amazon's AI Wallet: AWS, Coinbase, and Stripe building payment rails for AI agents — agents can buy APIs, web content, soon hotel bookings and merchant payments (CoinDesk, May 7)
MARA expected Q1 losses but investors focused on long-term AI growth strategy over Bitcoin volatility

📊 Key Trends & Signals This Week

Agent Orchestration is the #1 Theme: ruvnet/ruflo approaching 50K total stars (+7K this week) — multi-agent swarm intelligence and coordination is the hottest area in AI open source
Developer AI Agent Tooling Explosion: DeepSeek-TUI (+20K stars/week) and agent-skills show developer tooling for AI agents is the fastest-growing category
Local AI Research Agents Maturing: LearningCircuit/local-deep-research achieving ~95% on SimpleQA with local models (Qwen3.6-27B on a 3090) — privacy-first AI is viable
RAG Revolution: VectifyAI/PageIndex introducing vectorless, reasoning-based RAG — traditional vector database approaches being replaced by document indexing
AI Coding Access Democratization: 9router (+5K stars/week) connecting Claude Code, Codex, Cursor to 40+ free providers — unlimited AI coding for everyone
AI Agents in Finance: Anthropic launching financial-services reference architecture, plus AI-Trader and TradingAgents — AI agents in financial services going mainstream
Multimodal Agents Expanding: bytedance/UI-TARS-desktop connecting cutting-edge AI models with agent infrastructure for multimodal tasks
AI Agent Memory: agentmemory (#1 persistent memory solution) and ruflo's RAG integration — agent memory/long-term context is critical infrastructure
Bot Detection Arms Race: CloakBrowser (+5K stars/week) — stealth Chromium passing all bot detection, showing demand for automated web interaction
Enterprise AI Backend: InsForge giving AI coding agents full-stack backend (DB, auth, storage, compute, AI gateway) — end-to-end agent development platforms

🔑 Actionable Signals for Facorreia

AI Agent Memory is a solved-enough problem now (rohitg00/agentmemory trending) — worth evaluating for Norviq or second-brain app
Agent Orchestration (ruflo, UI-TARS) patterns could inform Loci's itinerary automation features
RAG Evolution (vectorless/PageIndex) is worth tracking for ObsidianBrain architecture decisions
Local Deep Research at 95% SimpleQA with local models — feasible approach for the Second Brain app's research features

Report generated: 2026-05-13T09:21:45Z Next refresh: Weekly cron Sources: arXiv API, GitHub Trending (browser), HuggingFace API, CoinDesk AI tag

ai-market-intelligence-2026-05-13