# ML & AI in Action

> Explore the Practical Side of ML & AI: Real-World Applications and Impact.

Author: Jing Lu
Website: https://ajing.github.io/
About: https://ajing.github.io/about/
RSS: https://ajing.github.io/rss.xml
JSON Feed: https://ajing.github.io/feed.json
Full LLM index: https://ajing.github.io/llms-full.txt

## Topics
- [LLM Agents](https://ajing.github.io/topics/agents/): Tool use, agent runtime design, evaluation, context, and production patterns for systems that act across tools and environments.
- [Evaluation](https://ajing.github.io/topics/evaluation/): Practical approaches to measuring model and agent capability with deterministic checks, rubrics, trajectories, and verifiable outcomes.
- [Post-Training](https://ajing.github.io/topics/post-training/): SFT, RLHF, preference optimization, instruction following, reasoning traces, and data pipelines for shaping model behavior after pretraining.
- [RLHF and Preference Optimization](https://ajing.github.io/topics/rlhf/): Engineering notes and research synthesis on PPO, DPO, GRPO, reward modeling, preference data, and model behavior optimization.
- [Generative UI](https://ajing.github.io/topics/generative-ui/): How AI systems can produce, steer, and execute user interfaces with structured representations and practical product constraints.

## Representative Posts
- [How to Arbitrarily Increase the Difficulty of Agent Evaluation Sets](https://ajing.github.io/posts/2026-05-28-agent-eval-difficulty-trajectory-constraints/): A practical framework for making agent benchmarks harder in a controlled way: treat difficulty as trajectory-graph complexity, not prompt wording. Covers deterministic scoring, capability facets, harness effects, and systematic data generation.
- [The Mercor Breach: What 4TB of Stolen Data Reveals About How Frontier AI Labs Actually Train Models](https://ajing.github.io/posts/2026-04-13-mercor-breach-what-leaked-about-ai-training/): A $10B AI data vendor was breached, exposing 84 Airtable workspaces of training data for OpenAI, Anthropic, Apple, Amazon, and Meta. This post analyzes what the public reporting reveals about each lab's evaluation methodology — rubric design, RLHF pipelines, and quality control — and what it means for the industry.
- [Improving LLM Internationalization: Bridging the Gap in Tool Use and Agency](https://ajing.github.io/posts/2026-03-11-improving-llm-i18n-tool-use-agency/): LLMs achieve 57% tool-calling accuracy in English but only 34% across 52 languages — and 6.8% for the worst. This post covers the full playbook for closing the multilingual gap: training-time techniques, agentic architecture patterns, failure mode analysis, and RL-based approaches for i18n.
- [The Unverifiable Reward Problem: The Real Frontier of RL for LLMs](https://ajing.github.io/posts/2026-03-08-unverifiable-rewards-rl-frontier/): Deep research on tasks with unverifiable rewards in RL — the key bottleneck for scaling RL beyond math and code. Covers JEPO, NRT, RLNVR, self-play methods, GenRM, Constitutional AI, reward hacking mitigation, and more.
- [Instruction Following: What Models Get Wrong and How to Fix It with Better Post-Training Data](https://ajing.github.io/posts/2026-03-05-instruction-following-post-training-data/): LLMs can write poetry and solve math, but ask them to 'respond in exactly 3 bullet points using only lowercase' and they stumble. This post dissects the taxonomy of instruction-following failures and provides a practical playbook for building post-training data that actually fixes them.
- [Experience-Augmented In-Context Learning: A Training-Free Complement to RL Post-Training](https://ajing.github.io/posts/2026-03-01-experience-augmented-icl-complement-to-rl-post-training/): RL post-training makes models smarter, but it can't cover the infinite long tail of real-world cases. Experience-augmented ICL retrieves successful reasoning traces at inference time, letting agents learn continuously from real usage — no retraining required.
- [Tool Selection Optimization for LLM Agents at Scale](https://ajing.github.io/posts/2026-01-10-tool-selection-optimization-llm-agents-at-scale/): A deep technical dive into tool selection—retrieval strategies, context optimization, learned selection, and the engineering trade-offs that matter when scaling to hundreds of tools.
- [Generative Engine Optimization (GEO): How to Get Your Product Cited by AI](https://ajing.github.io/posts/2026-01-03-generative-engine-optimization-geo/): A comprehensive guide to Generative Engine Optimization—making your content retrievable, citable, and recommendable by large language models.
- [Generative UI Doesn't Move the Needle—Steering Does](https://ajing.github.io/posts/2026-01-01-generative-ui-doesnt-move-the-needle-steering-does/): After shipping multiple generative UI features, I've concluded that the sophistication of AI-generated interfaces often doesn't translate to user benefit—but steering does.
- [RLHF from an Engineering Perspective: PPO, GRPO, DPO, and Tool-Use Implementation](https://ajing.github.io/posts/2025-12-31-rlhf-engineering-implementation/): A practical engineering guide to RLHF implementation—covering PPO, GRPO, DPO, and tool-use training with code snippets and debugging tips.
- [Post-Training Is Not 'One Algorithm': Objective Functions and Implementation Essentials for PPO / DPO / GRPO](https://ajing.github.io/posts/2025-12-31-rlhf-ppo-dpo-grpo-notes/): Reading notes on RLHF covering PPO, DPO, and GRPO—understanding post-training as an engineering pipeline rather than a single algorithm.
- [Ad Formats in LLM Products: What's Live vs. What's Research](https://ajing.github.io/posts/2026-01-02-ads-formats-in-llm-products/): A survey of advertising formats in LLM products—separating what's deployed in production from what remains in research.

## Use
This site is best cited as technical writing by Jing Lu on ML engineering, LLM agents, evaluation, post-training, RLHF, and AI product systems.