In another reckless expansionist provocation, the Philippine government has delimited the so-called "baselines of the territorial sea" of China's Huangyan Dao, a move that flagrantly violates China's sovereignty and international law and is doomed to fail.

RAG's "more precision means more latency" dilemma, tackled by using topics as a compass 🧭 Title: MCompassRAG: Topic Metadata as a Semantic Compass for Paragraph-Level Retrieval URL: 🧭 Overview A metadata-guided retrieval framework that uses topic-level signals as a "semantic compass" to select relevant evidence at the paragraph level. It aims to improve precision and efficiency at the same time. ❓ Challenges Solved RAG faces a precision-vs-efficiency trade-off. ・Fine-grained chunks raise precision but increase candidates, latency, and cost ・Larger chunks reduce candidates but introduce semantic noise from mixed topics This is acute in deep-research tasks needing fast, precise retrieval over large datasets. 💡 Methodology & Proposed Approach ・It enriches chunk representations with topic metadata within the same embedding space ・It uses LLM-teacher distillation to train a lightweight retriever ・This enables topic-aware retrieval with no additional LLM calls at inference time The core is combining metadata with dense embeddings and distilling into a lightweight retriever. 📊 Experimental Results ・Information efficiency: 8.24% average improvement across six benchmarks ・Latency: over 5x lower than the strongest efficient RAG baselines ・Code is available in a public repository It achieves this precision-and-speed balance without extra LLM calls at inference. #RAG# #Retrieval#

AI systems now write research papers autonomously — yet phantom references, method-code misalignment, and unreproducible scores have become endemic failures undermining scientific integrity. Title: Science One Framework: A verifiable autonomous research framework via Chain-of-Evidence Google Cloud's Science One Framework treats verifiability as a first-class architectural constraint through the Chain-of-Evidence principle, solving the trustworthiness crisis in autonomous AI research at its root. 🔍 Highlight 1 — Chain-of-Evidence (CoE): two foundational principles Completeness: every claim carries a recorded evidence chain. Correctness: each chain genuinely supports its claim. These two principles eliminate phantom references entirely — baseline systems showed rates up to 21% — while achieving best-in-class method-code alignment across all evaluated systems. The key difference from prior work: evidence chains are constructed at claim-generation time, not retrofitted as a post-hoc check. 🏗 Highlight 2 — Three-module architecture Problem Investigator builds citation graphs from up to 100 full-text PDFs via Semantic Scholar API, grounding every reference in retrieved data rather than model memory. Discovery Engine explores parallel solution branches while keeping immutable records of all raw evaluator outputs. Paper Writer and Claim Verifier binds every factual claim to specific workspace artifacts and conservatively reconciles misalignments rather than deleting them — preserving scientific transparency. 🏆 Highlight 3 — MLE-Bench and Parameter-Golf results Across five Kaggle competitions covering medical imaging, fine-grained recognition, and 3D perception: two Gold Medals and two Silver Medals. Won the 3D Object Detection task where every baseline system failed completely. On Parameter-Golf — a live LLM training competition under strict hardware and file-size constraints — achieved state-of-the-art as of April 27, 2026, while baselines could not produce valid submissions at all. Rigor and capability don't trade off. Science One outperformed five state-of-the-art systems including AI Scientist v2, AutoResearchClaw, and DeepScientist, setting a new standard for verifiable autonomous research. #AIResearch# #AutonomousScience#

AI can now remember, update, and forget facts without any external retrieval. A new paradigm for native memory in foundation models has just been published. Title: Metis: Memory Foundation Model 🔍 Overview Current external memory systems like RAG face three fundamental limitations: decoupling from the model backbone, inability to propagate gradients through discrete memory operations, and added inference latency. Metis integrates memory natively into Transformer blocks — analogous to how Chain-of-Thought became intrinsic to LLMs — eliminating external module dependency while enabling learned, end-to-end memory behavior. 🛠 Problem and Approach Two core components are introduced. The Local Memory Block maintains a dense memory network updated across inference steps via exponential moving average, with learned importance scoring and Top-ρ token selection. The Hyper Memory Block uses static learned parameters to enable memory transformation through the forward pass. Four memory operations — Remember, Update, Forget, and Reflect — are executed purely through forward computation without any background gradient updates. 📊 Experimental Results On the MemOps benchmark (no-context setting), Metis-27B achieves 24.76% average performance: · Baseline Qwen3.5-27B (no context): 1.69% · Test-time training Temp-LoRA-27B: 9.70% · Parametric memory δ-Mem: 4.38% On the Metis internal test set, the model reaches 73.77% average with Reflect (multi-hop reasoning) at 93.44%. It outperforms all no-context baselines across model sizes and tasks. 💡 Practical Significance Native memory avoids the retrieval, ranking, and prefilling overhead of RAG through parallelizable computation, keeping inference latency low. Domain adaptation via post-training is supported. Model checkpoints and code are publicly available on GitHub (MemTensor/Metis) and HuggingFace. #LLM# #AIAgent#

💰 Don't size caches for peak—let them stretch and shrink with demand to cut cost. Google applies the classic ski rental problem to a production database. Title: Optimizing cloud economics with linear elastic caching URL: 📦 Overview Linear elastic caching treats memory footprint as a variable cost that integrates over time, dynamically growing and shrinking cache size to match the workload. 🎯 The problem Cloud memory is expensive (serverless can charge up to $3/day per GiB). Fixed-size caches hit a "Goldilocks dilemma": too small hurts performance, too large wastes thousands on idle memory during low demand. 🎿 Method Each page faces a choice: "rent" (keep in RAM, paying continuous memory cost) or "buy the miss" (evict, risking latency/I/O later). A ski rental algorithm sets each page's TTL. The key result: eviction policy and rental duration can be optimized separately. 🌲 Implementation For Spanner, a lightweight shallow decision tree (compilable to C++) predicts the optimal TTL from data size, miss cost, and operation type—no heavy inference in the cache path. 📉 Results In production on Spanner: memory down 15.5%, misses up only 5.5%, TCO down ~5%, I/O impact a mere 0.5%. On public traces it consistently beat fixed-size (GDSF) baselines. #CloudComputing# #Algorithms#

🔎 A provocative paper: "agentic search needs no embeddings or vector DB—just grep the raw corpus directly." Title: Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction URL: ❓ What is Direct Corpus Interaction (DCI)? 💡 A retrieval paradigm where the agent explores the raw corpus directly via grep, find, and shell commands—no embedding model, vector index, or retrieval API. It needs no offline indexing and adapts naturally to evolving local corpora. ❓ Why aren't conventional retrievers enough? 💡 Sparse or dense, they compress the corpus through a fixed similarity interface into top-k before reasoning begins. That makes exact lexical matching, combining weak clues, and checking local context hard—and evidence filtered out early can't be recovered no matter how strong the downstream reasoning. That's fatal for multi-step agent search. ❓ Can it really win without a retriever? 💡 Yes. On BrowseComp-Plus, swapping the retriever for DCI on the same Sonnet 4.6 lifts accuracy 69.0 → 80.0% (+11.0) while cutting cost 29.4%. Multi-hop QA averages 83.0 (+30.7 over the strongest baseline); IR ranking hits 68.5 NDCG@10 (+21.5). Even the lightweight GPT-5.4 nano version beats many baselines. ❓ What's actually driving it? 💡 The authors call it "retrieval interface resolution." Trajectory analysis shows the edge comes less from surfacing more gold documents and more from high-resolution local search and verification after reaching them. #AIAgents# #RAG#

# Practices for Embedding AI Agents in Software # Prompt Registry / Prompt Artifact 🎯 The Hook A one-word prompt change shipped inside a large commit caused a production incident. No version history, no rollback path, no way to tell which prompt produced which decision. Sound familiar? 🔥 The Problem In LLM-based systems, a small prompt edit can dramatically alter model behavior. When prompts live as string literals scattered across application code, change tracking is impossible, regression testing doesn't exist, rollback requires a full code deploy, and auditors can't determine which prompt version drove a given decision. This is especially problematic in regulated domains where accountability demands a clear link between prompts and outcomes. 💡 The Pattern Treat prompts as first-class versioned artifacts with the same rigor as application code: version control, peer review, automated regression testing via evaluation harnesses, and staged deployment. Roll out prompt changes through canary releases at 5-10% traffic, automatically rolling back on quality degradation. Record the prompt ID and version in every execution trace so post-incident audits can pinpoint exactly which prompt was active. Track prompt-model compatibility to identify which prompts need re-evaluation when the underlying model is updated. ✅ When to Use Use when: - Prompt change history and runtime version tracking are required (regulated industries, quality management) - Prompts are referenced from multiple places and need centralized management - You want A/B testing or gradual rollout capabilities for prompt changes Don't use when: - There are only one or two prompts with low change frequency in a personal project or PoC - Output quality variation is acceptable in an exploratory context ⚠️ Pitfalls - Template variables populated with user input need injection protection through escaping or sanitization - Over-relying on an external registry API for prompt resolution introduces availability risk. Consider fetch-at-startup with local caching - Store prompts in line-oriented formats like YAML or Markdown. A giant single-line JSON makes diff review nearly impossible 🔧 Implementation Approach - Store prompts in line-oriented formats (YAML/Markdown) with template content, variable definitions, model compatibility constraints, and evaluation baselines in a single file - Resolve prompts at runtime by ID from the registry, incorporating probabilistic canary routing to serve canary versions to a subset of traffic - Record prompt ID and version in every execution trace, enabling post-incident audits to pinpoint which prompt drove each decision - Integrate evaluation harnesses into CI to automatically run regression tests on prompt changes, then roll out via canary releases with automatic rollback on quality degradation #AIAgents# #SoftwareArchitecture#

🎼 Solve hard tasks that mix text, image, audio, and video by decomposing them across specialized sub-agents running in parallel. This work shows a well-matched team beats one giant monolithic model. Title: Orchestra-o1: Omnimodal Agent Orchestration URL: 💡 Overview Orchestra-o1 is a hierarchical agent framework that separates high-level orchestration from low-level tool execution to handle tasks where multiple modalities coexist. It specializes sub-agents per modality and runs independent sub-tasks in parallel for efficiency. ⚠️ The problem Existing orchestration frameworks handle only limited modalities and fail to generalize to scenarios where text, image, audio, and video coexist and interact at once. 🛠 Approach ・Represent each backend with a skill vector plus a cost-latency profile, and select via cost-aware matching ・Assign perception tools (image/audio/video analysis) and action tools (search, page visit, code execution) ・Build a latent dependency graph over sub-goals and run independent tasks in parallel ・Train with DA-GRPO: score step-level decisions, not just final answers, using a multi-dimensional rubric that weights decision quality at 0.6 📊 Results (their OmniGAIA benchmark) ・Orchestra-o1-GPT-5 hits 72.8%, beating the second-best Gemini-3-Pro by 10.3 points ・The open-source Orchestra-o1-8B (from Qwen3-8B) reaches 30.0%, best among open omnimodal agents ・It reaches 72.8% at cost 341.6, cheaper and far more accurate than weaker baselines ・By difficulty: Easy 80.3%, Medium 75.0%, Hard 56.4% #AIAgents# #Multimodal#

🪑 Insert an object into an image while specifying its exact 3D orientation and position. DIRECT solves the 3D-pose control that text leaves ambiguous and parameters struggle with, by decomposing visual proxies. Title: Direct 3D-Aware Object Insertion via Decomposed Visual Proxies URL: 📝 Overview DIRECT is a diffusion-based method that inserts a reference object into an image with explicit control over its 3D pose and position. It decomposes the insertion condition into geometry, appearance, and context, injected through independent pathways. ❓ Challenges Solved Existing insertion methods formulate the task as 2D inpainting and can't control 3D pose. Text guidance is spatially ambiguous, and parametric 3D methods can't translate abstract parameters into correct geometric projections. 💡 Methodology & Proposed Approach ・A user-manipulated 3D proxy rendered at the target pose provides geometry guidance ・Appearance (the reference's high-fidelity look) and context (background semantics) are injected independently via separate LoRA adapters and positional embeddings to avoid feature entanglement ・TRELLIS lifts the image into a coarse 3D shape, refined with VGGT and 3D Gaussian Splatting ・Built on FLUX.1-Fill, it uses shape-decomposed mask augmentation and progressive-resolution training to avoid overfitting 🎯 Use Cases It fits virtual staging, e-commerce product photography, creative work needing precise spatial control, and photorealistic AR/VR content generation. 📊 Experimental Results ・On the FLUX backbone it reaches PSNR 23.09, LPIPS 0.147, and matching error 17.8, beating baselines on all metrics ・It stays stable across large 0-180 degree pose changes and preserves fine details even under 3D-reconstruction degradation ・Hybrid-data training raised CLIP-I from 0.904 to 0.943 ・For symmetric object orientation, RGB geometry guidance outperformed normal maps #3DGeneration# #ImageEditing#