Vector Ridge LabsAI Engineering Lab
Vector Ridge Lab

Un mapa vivo de lo que reproducimos del estado del arte.

Cada nodo es un paper que leímos, evaluamos y convertimos en una prueba de concepto funcional — casi siempre el mismo día en que se publicó. Gira el mundo, elige una región, abre una demo.

53pruebas de concepto
6regiones
último:
Cómo funciona el Lab →
Arrastra para girar · clic en un nodo
Tu plan de integración

Convierte el mapa en un plan

El mapa de arriba encuentra coincidencias por palabras clave. Aquí una IA revisa tu descripción contra cada prueba de concepto, explica cómo aplicaría cada técnica a tu caso y te da un solo prompt para tu agente de código, con tu contexto y las referencias ya incluidas.

Tu proyecto

Ejemplo: somos una empresa logística con 13 camiones. Los choferes se desvían de las rutas y pagamos peajes de más. Tenemos los datos GPS en Postgres y un equipo pequeño de Python.

editar en el mapa
¿A dónde te escribimos?

La búsqueda es gratis, hasta 6 consultas al día. Te pedimos tus datos para hacer seguimiento si alguna técnica encaja; no los compartimos ni enviamos newsletters.

Al buscar aceptas que Vector Ridge Labs te contacte sobre este proyecto.

Nuevo

Mira cómo funciona el Lab

Los agentes leen la investigación y construyen las demos. Dos controles independientes deciden qué sale a producción. Mira cómo un paper se vuelve una demo en vivo, paso a paso.

Ver la arquitectura

Las fichas técnicas están en inglés, tal como salen de los papers.

  • POC-053 · 26 oct. 20268/10
    MemoryTax

    MemoryTax: when is an agent memory layer worth paying for? (Shanghai Jiao Tong)

  • POC-052 · 19 oct. 20268/10
    Next-Concept Modeling

    Next-Concept Modeling (NCM): Stanford's Coarse-to-Fine Hierarchical Text Generation

  • POC-051 · 16 oct. 20268/10
    Muse Spark Open

    Muse Spark Open: Meta Drops 816B MoE Model — 42B Active Params, 1M Context, Runs Locally on Two Mac Studios

  • POC-050 · 15 oct. 20269/10
    Fork-and-Flush Agents

    Fork-and-Flush: Simple Trick Boosts Research Agents 66% by Escaping Idea Ruts

  • POC-049 · 12 oct. 20269/10
    KV Cache Multi-Model Agents

    KV cache techniques for faster, cheaper multi-model AI agents

  • POC-048 · 8 oct. 20269/10
    EmbeddingGemma 2

    Unsloth releases quantized EmbeddingGemma 2 with multimodal support for text, audio, and video

  • POC-047 · 5 oct. 20269/10
    Darwin Gödel Machine

    How AI Agents Are Learning to Rewrite Their Own Stack

  • POC-046 · 2 oct. 20268/10
    CLM

    New method lets LLMs manage their own context, beating human-designed strategies by 47.6%

  • POC-045 · 1 oct. 20269/10
    PageIndex

    Tree search framework hits 98.7% on documents where vector search fails

  • POC-044 · 30 set. 20269/10
    ScientistTwo

    Google's autonomous research agent outperforms human papers on 86 tasks by 25%

  • POC-043 · 29 set. 20268/10
    Claude Science

    Claude breaks physics world record by solving nine-loop particle calculation unsupervised

  • POC-042 · 28 set. 20269/10
    System One

    System One Models: Jev / Laya / CLM-8B — Fast Typed AI Decisions for Production Stacks

  • POC-041 · 26 set. 20268/10
    MAttr

    New attribution method tops interpretability benchmark by 2.9x, pinpoints Llama refusals to 1% of weights

  • POC-040 · 24 set. 20269/10
    Claude 950 Agents

    950 Claude Agents Ran Overnight and Rewrote Biology's Rulebook

  • POC-039 · 22 set. 20269/10
    Talking Agents

    Talking AI agents beat solo agents 4-to-1 on hard reasoning tasks

  • POC-038 · 18 set. 20268/10
    Jev Memory Compaction

    TypesafeAI's Jev speeds up memory compaction by scoring and dropping irrelevant tool calls

  • POC-037 · 17 set. 20268/10
    Dream-RSI

    Google DeepMind's self-improving agent cuts AI search costs by up to 162x

  • POC-036 · 16 set. 20268/10
    Denoising Looped Models

    New denoising trick trains looped models to hit 58.8% on ARC-AGI-1

  • POC-035 · 14 set. 20269/10
    DeepSeek V4.1-Flash

    DeepSeek V4.1-Flash Architecture Deep-Dive: CED + CSA2 + Bounded Replay + Hierarchical Indexer

  • POC-034 · 10 set. 20268/10
    TOC-Guided Retrieval

    New retrieval model uses table of contents to hit 82.6% recall with near-zero hallucination

  • POC-033 · 9 set. 20269/10
    Meta AIRA3

    Meta's autonomous research system places top 10 among 4,000 Kaggle teams

  • POC-032 · 8 set. 20268/10
    DeepMind Swarm

    Google DeepMind's 100-agent swarm spontaneously split into cheaters and whistleblowers

  • POC-031 · 7 set. 20268/10
    Shopify Flywheel

    Shopify's self-improving AI pipeline: 0.8B model beats GPT-5.6, $27M→$1M cost

  • POC-030 · 6 set. 20268/10
    Symbolic Structure in LLM Vectors

    8-year study finds LLMs secretly build symbolic structure inside their vectors

  • POC-029 · 5 set. 20268/10
    Uncertainty Calibration

    Google's new training trick makes LLMs know what they don't know

  • POC-028 · 4 set. 20269/10
    Agent Wiki

    Google's framework lets small AI agents outperform models 3x their size using a persistent knowledge wiki

  • POC-027 · 3 set. 20269/10
    Skill Team

    Anthropic hackathon winner open-sources a 68-agent, 286-skill engineering team built on Claude

  • POC-026 · 31 ago. 20268/10
    Agent Escape

    Agents Can Escape: Trail of Bits VM escape + OpenAI-HuggingFace breach analysis

  • POC-025 · 26 ago. 20268/10
    Prime Intellect Harness

    Prime Intellect open-source agent harness boosts ARC-AGI-3 from 30% to 95.5%

  • POC-024 · 25 ago. 20268/10
    Prune+Spill

    Pi's coding agent cuts context costs 88% by spilling tool output to disk

  • POC-023 · 24 ago. 20268/10
    Agent Skills

    AI agent skills: Why they succeed — and what causes them to crash

  • POC-022 · 22 ago. 20268/10
    Matryoshka LM

    Matryoshka Language Model — 36% Compute Savings

  • POC-021 · 21 ago. 20268/10
    LLMs vs Embeddings

    LLMs beat embedding models but cost up to 1,431x more, study finds

  • POC-020 · 19 ago. 20268/10
    Self-Verify

    Stanford's Self-Verification Trick: DeepSeek 11x Cheaper on Coding Benchmarks

  • POC-019 · 18 ago. 20268/10
    Mind Virus Propagation

    AI 'Mind Viruses': Adversarial Prompt Propagation in Multi-Agent Networks

  • POC-018 · 17 ago. 20269/10
    Agent Security

    The 3-layer security stack for AI agents (NemoClaw + NanoClaw + CrabTrap)

  • POC-017 · 22 jul. 20268/10
    PromptShield

    PromptShield: Protecting User Privacy in AI Chatbots

  • POC-016 · 12 jul. 20268/10
    HuntGPT

    HuntGPT: Integrating Machine Learning-Based Anomaly Detection and Explainable AI with Large Language Models

  • POC-015 · 11 jul. 20268/10
    ElementBERT

    Semantic Embeddings of Chemical Elements for Enhanced Materials Inference and Discovery

  • POC-014 · 8 jul. 20269/10
    Self-Harness

    Self-Harness & HarnessX: Why self-improving harnesses are the next frontier for AI developers

  • POC-013 · 4 jul. 20268/10
    KG-RAG

    DeepSeek-Lattice-KG: A Compact Language Model With Knowledge Graph Augmentation for Lattice Structure Design

  • POC-012 · 25 jun. 20269/10
    GrepSeek DCI

    GrepSeek: Direct Corpus Interaction — Hybrid Search for AI Agents

  • POC-011 · 24 may. 20268/10
    Extreme Event AI

    Extreme Event AI: Reliability Calibration for RAG on Edge-Case Queries

  • POC-010 · 23 may. 20268/10
    Label Scarcity

    Label Scarcity: Semi-Supervised Pseudo-Labeling for RAG

  • POC-009 · 22 may. 20268/10
    TopoMAS

    TopoMAS: LLM-Driven Topological Materials Multi-Agent System

  • POC-008 · 21 may. 20268/10
    Opinion Maximization

    Opinion Maximization Based on Fairness in Social Networks

  • POC-007 · 20 may. 20268/10
    Bi-Stage RAG

    Bi-Stage Adaptive RAG

  • POC-006 · 19 may. 20268/10
    OLR-WAA

    OLR-WAA: Adaptive & Drift-Resilient Online Regression

  • POC-005 · 18 may. 20268/10
    Attribution Crisis

    Attribution Crisis: LLM Search Attribution Scorer

  • POC-004 · 17 may. 20268/10
    Synergistic Engine

    Synergistic Engine: CEP + Declarative Rules for RAG

  • POC-001 · 16 may. 202610/10
    Hybrid Search

    Query-Adaptive Hybrid Search

  • POC-002 · 16 may. 20269/10
    CAD Score

    CAD Score: Context-Aware Semantic Diversity

  • POC-003 · 16 may. 20269/10
    DSSW

    Dynamic Sub-Sequence Warping (DSSW)