Comprende la IA.
Aprende con cada experimento.
Un espacio de lectura y experimentación para ingenieros e investigadores de IA. Estudia las ideas, sigue una ruta y ponlas a prueba.
Una ruta completa
Leer → implementar → evaluar
Investigación de frontera
Ocho lecturas seleccionadas: de las ideas a los experimentos.
Lectura destacadaRobot recovery: when retrying helps, and when it hides failure
Posted
Read Backtrack, DexPort and ViT-H through a practical lens: frozen policies, action labels, executable motion and bounded recovery. Includes a local state-machine experiment.
Notas de investigaciónSelf-improving agents: reward gains are not capability gains
Posted
A research reading guide and local experiment on adaptive harnesses, reward collapse, frozen-policy comparisons and held-out evaluation.
Notas de investigaciónFrom LLM Calls to Agent Harnesses
Posted
An English reading brief and hands-on tutorial on agent runtimes: context, tool permissions, checkpoints, bounded execution and reproducible failure tests.
Notas de investigaciónTest-time compute: search, verification and budgets
Posted
Prerequisites: probability, evaluation and inference cost. Learn when more attempts help and when a verifier becomes the bottleneck.
Notas de investigaciónSelective state spaces: Mamba and sequence memory
Posted
Prerequisites: linear algebra, recurrence and attention. Compare compressed state with explicit token retrieval.
Notas de investigaciónDiffusion language models: masking and iterative decoding
Posted
Prerequisites: cross-entropy, Transformers and generative modeling. Understand LLaDA-style masked generation without equating parallelism with speed.
Análisis de artículosDeepSeek-R1: rewards, reasoning and what the evidence establishes
Posted
Study R1-Zero and R1 separately, then build a controlled test of reasoning improvement. This analysis is pinned to arXiv v1.
Análisis de artículosLLM-as-a-judge: evaluate the evaluator
Posted
Separate answer preference from correctness and reveal how order effects can alter a reported win rate.
Biblioteca de aprendizaje
Elige una ruta de aprendizaje.
Plan de estudios de IA / LM
De los fundamentos a los modelos de lenguaje.
Aprendizaje por refuerzo
Funciones de valor, políticas y razonamiento.
IA física
Percepción, control y aprendizaje robótico.
Biblioteca de artículos
Mecanismos, derivaciones y evidencia.
Últimos tutoriales
Novedades de la biblioteca
Write Multi-Head Attention from scratch
Trace Q, K and V, split and merge heads, then prove your causal mask works.
Leer tutorial ↗Softmax that stays stable
Derive the gradient, work through a small example and avoid overflow.
Leer tutorial ↗K-Means beyond the happy path
Implement assignment and updates; handle empty clusters and convergence.
Leer tutorial ↗Adam, one update at a time
Understand persistent moments, bias correction and the first parameter update.
Leer tutorial ↗Beam Search without hidden mistakes
Keep global top-k candidates, track EOS and compare against greedy decoding.
Leer tutorial ↗Experimenta localmente
Abre una herramienta y pon una idea a prueba.
Las herramientas procesan los datos en tu navegador. Sin claves API ni cargas de entradas.
Entrevistas y proyectos
Practica con evidencia, no con respuestas memorizadas.
Práctica de entrevistas
Preguntas, respuestas y razonamiento.
Proyectos y código
Implementaciones y pruebas de aceptación.
Práctica reciente
Ejercicios de entrevista y proyectos prácticos
Handwrite Beam Search, then defend every choice
Practice global top-k selection, EOS handling and length bias with a worked example and reference code.
Ver ejercicio ↗INTERVIEW / OPTIMIZATIONDerive Adam and trace the first update
A paper-and-pencil drill in moving averages and bias correction, followed by a tested implementation.
Ver ejercicio ↗PROJECT / TRANSFORMERSProve your KV cache is correct
Build causal attention and compare full-sequence outputs with incremental decoding using numerical invariants.
Ver ejercicio ↗PROJECT / RAGBuild retrieval with a measured baseline
Measure retrieval accuracy, answerable recall and correct abstention separately; inspect the failure cases.
Ver ejercicio ↗