Bojian Li

🧠 AI Rights Historical Wrongs Researchers at Stanford and Princeton have fine-tuned a large language model to uncover racially discriminatory housing clauses buried in millions of historical property deeds — accelerating a process that could have taken a decade by hand. 📍 The problem: Many property deeds in Northern California — especially in Santa Clara County — still contain illegal racial covenants from as far back as the 1850s, banning people of color from owning or living in certain homes. 🔎 The solution: The team used OCR to digitize 5.2 million deed pages, then fine-tuned Mistral-7B via LoRA on a hand-labeled dataset of discriminatory phrases. The model learned to identify harmful language in context, avoiding false positives like the surnames "Black" or "White." 📊 Results: Detected 24,500 lots with racial clauses (~25% of homes in 1950) Found that 10 developers were responsible for 1/3 of all discriminatory clauses Model reviewed all pages in 6 days at a cost of $258 Equivalent manual effort: 10 years, $1.4M 💥 This work doesn't just use LLMs to build the future — it uses them to understand and repair the past. The open-sourced model now enables other U.S. counties to follow suit. #AI #MachineLearning #LLM #ResponsibleAI #Stanford #Princeton #HistoryMeetsAI #TechForGood #EthicalAI #SocialImpact #PropertyTech https://lnkd.in/dtec-zbx

4

Do we get some a free DGX Spark depending on the number of companies we have worked in our career that showed up today during Jensen Huang's NVIDIA GTC keynote? 🤣

We (w/ Shansan Gong, Ruixiang ZHANG, Huangjie Zheng, Jiatao Gu, Navdeep Jaitly, Lingpeng Kong) released a family of 7B diffusion language models, DiffuCoder, that specializes on code generation, with a focus on understanding and improving masked diffusion models. A core analysis of DiffuCoder is the autoregressiveness (AR-ness) score, a novel metric that quantifies the causal patterns in decoding, revealing how diffusion models break from strict left-to-right generation for more flexible, non-linear code planning. Recent advances in autoregressive (AR) models dominate code generation, but diffusion-based LLMs (dLLMs) like DiffuCoder offer a promising alternative, especially for complex programming tasks. DiffuCoder explores how these models decode differently—showing less global AR-ness in code tasks compared to math—and how temperature affects both token selection and generation order, unlike traditional AR models. We also introduce coupled-GRPO, a post-training RL method with a coupled-sampling scheme, to reduce performance drops during accelerated decoding, boosting parallelism and efficiency. We use a self-improvement pipeline that leverages AR-ness analysis, coupled-GRPO optimization, and evaluation on benchmarks like AceCode-89k to refine decoding strategies. This approach enables DiffuCoder to navigate diverse code generation pathways and enhance performance with modest computational overhead. Looking ahead, we aim to further leverage Reinforcement Learning to steer code generation through these decoding patterns, with the discrete nature of AR-ness scores providing a foundation for search-based strategies—ideal for the sparse rewards of optimizing complex code structures. Check out our full paper and code for a deeper dive! Paper: https://lnkd.in/gVWU3BDJ Code: https://lnkd.in/gmXTZ_6n Models: https://lnkd.in/gTcKCDr9 #MachineLearning #AI #CodeGeneration #DiffusionModels #NLP

220

5 Comments

Evaluating Large Language Models (LLMs) is no longer just about accuracy scores. As LLMs move into production and power real-world applications, evaluation must balance scale, nuance, and trust. Automatic Evaluation is the first layer. It relies on predefined metrics and usually a ground-truth reference, making it fast and cost-effective for large datasets and objective tasks. Metrics like BLEU (n-gram overlap for translation), ROUGE (content recall for summarization), exact match/accuracy (factual Q&A), and semantic similarity (embedding-based meaning comparison) help quantify model performance where correctness is clearly defined. LLM-as-a-Judge (LLMaaJ) goes a step further. Here, a separate and often stronger LLM evaluates model outputs against qualitative criteria. This approach captures aspects traditional metrics miss—such as helpfulness, coherence, relevance, tone, and completeness—and works even without reference answers. It is especially powerful for RAG systems, where the judge can assess both retrieval quality (context recall/precision) and generation quality (faithfulness and answer relevance). Custom metrics also allow evaluation aligned to domain or brand-specific requirements. Human-in-the-Loop (HITL) review remains critical. For highly subjective judgments, complex edge cases, and final production validation, human reviewers provide the gold standard. This is essential for assessing creativity, empathy, cultural fit, and domain-critical accuracy, especially in regulated areas like healthcare or legal, where expert curation helps uncover failure modes that automated systems cannot. In practice, strong LLM evaluation is layered—automatic metrics for scale, LLM-based judgment for depth, and human review for confidence and accountability. #LLMEvaluation #GenAI #AIEngineering #AgenticAI #RAG #MLOps #ResponsibleAI #AIQuality

6

Reinforcement Pre-Training (Microsoft, June 2025) Paper: https://lnkd.in/eY4PCpUi Abstract: "In this work, we introduce Reinforcement Pre-Training (RPT) as a new scaling paradigm for large language models and reinforcement learning (RL). Specifically, we reframe next-token prediction as a reasoning task trained using RL, where it receives verifiable rewards for correctly predicting the next token for a given context. RPT offers a scalable method to leverage vast amounts of text data for general-purpose RL, rather than relying on domain-specific annotated answers. By incentivizing the capability of next-token reasoning, RPT significantly improves the language modeling accuracy of predicting the next tokens. Moreover, RPT provides a strong pre-trained foundation for further reinforcement fine-tuning. The scaling curves show that increased training compute consistently improves the next-token prediction accuracy. The results position RPT as an effective and promising scaling paradigm to advance language model pre-training."

7

1 Comment

Google Research Director Denny Zhou, who founded the LLM Reasoning Team at DeepMind, recently gave a great talk at Stanford’s CS25 class: Large Language Model Reasoning. This was one of the most intuitive talks about reasoning model training and application. He outlined new directions for training models to handle questions where answers aren’t easily verifiable. My team is currently working on an enterprise LLM project focused on using reasoning to extract relevant context and generate responses for unverifiable cases. It’s a tough but promising area, and it’s an interesting time to be building. [Link to Denny Zhou's talk: https://lnkd.in/gdJ6Mzi8]

1,026

8 Comments

Want to truly understand how AI agents work? Build them yourself. I found a repository that teaches you exactly that—without LangChain or any abstractions. It focuses on raw code and real understanding. The course consists of 9 chapters that build on each other: - Loading local LLMs - Prompt engineering - Function calling (tools) - Persistent memory - ReAct reasoning loops Here's why this matters: Frameworks are beneficial until something breaks. When that happens, you're left debugging code you never fully understood. This repository addresses that issue. Start by building everything from scratch. Then, when you later use LangChain or CrewAI, you'll have a clear understanding of what's happening underneath. Check it out here: https://lnkd.in/gy_hVF6r

20