Editing and Proofreading Skills

New! We’ve published a new set of automated evaluations and benchmarks for RAG - a critical component of Gen AI used by most successful customers today. Sweet. Retrieval-Augmented Generation lets you take general-purpose foundation models - like those from Anthropic, Meta, and Mistral - and “ground” their responses in specific target areas or domains using information which the models haven’t seen before (maybe confidential, private info, new or real-time data, etc). This lets gen AI apps generate responses which are targeted to that domain with better accuracy, context, reasoning, and depth of knowledge than the model provides off the shelf. In this new paper, we describe a way to evaluate task-specific RAG approaches such that they can be benchmarked and compared against real-world uses, automatically. It’s an entirely novel approach, and one we think will help customers tune and improve their AI apps much more quickly, and efficiently. Driving up accuracy, while driving down the time it takes to build a reliable, coherent system. 🔎 The evaluation is tailored to a particular knowledge domain or subject area. For example, the paper describes tasks related to DevOps troubleshooting, scientific research (ArXiv abstracts), technical Q&A (StackExchange), and financial reporting (SEC filings). 📝 Each task is defined by a specific corpus of documents relevant to that domain. The evaluation questions are generated from and grounded in this corpus. 📊 The evaluation assesses the RAG system's ability to perform specific functions within that domain, such as answering questions, solving problems, or providing relevant information based on the given corpus. 🌎 The tasks are designed to mirror real-world scenarios and questions that might be encountered when using a RAG system in practical applications within that domain. 🔬 Unlike general language model benchmarks, these task-specific evaluations focus on the RAG system's performance in retrieving and applying information from the given corpus to answer domain-specific questions. ✍️ The approach allows for creating evaluations for any task that can be defined by a corpus of relevant documents, making it adaptable to a wide range of specific use cases and industries. Really interesting work from the Amazon science team, and a new totem of evaluation for customers choosing and tuning their RAG systems. Very cool. Paper linked below.

Context and Relevance. Two words that decide whether your story lands or drifts away like smoke. Everyone is a storyteller today. Leaders, marketers, founders, influencers. The world is flooded with narratives. But here’s the problem: most lose the plot. A story without context is noise. A story without relevance is nostalgia at best, irrelevance at worst. ⇢ A storyteller who ignores the listener’s world is only speaking to themselves. ⇢ A product designer who forgets the user’s need is only decorating, not designing. ⇢ A leader who misses the moment, the pain, the truth of their team is only performing, not leading. We love the glamour of words, the thrill of visuals, the romance of once upon a time. But if the story does not fit the frame of today’s reality or tomorrow’s aspiration, it will not stick. It is Monday morning. Your people are not starting from the same place. Some have already sprinted through their inbox. Others arrive carrying a weekend that never really gave rest. That gap is not a failure of effort. It is the human reality leaders walk into. And this is where leadership gets tested. Not in the size of your vision or the polish of your story, but in a simple pause that asks: Where are my people right now, and what matters most in this moment? That pause is context. That adjustment is relevance. Think of the last pitch that bored you. The last vision you rolled your eyes at. The last product that solved a problem nobody had. I have seen what happens when context and relevance go missing: ⇢ A leader speaks passionately about the company’s future while the team is drowning in today’s chaos. The vision is strong, but without context it floats above their heads. ⇢ A manager offers carefully worded feedback but never ties it to the individual’s reality. The intention is good, yet without relevance it does not land. ⇢ A consultant presents a smart, structured framework but never touches the client’s burning pain. The work is solid, but without connection it does not create movement. None of these moments come from bad leadership. They arise from missing the anchors. Context provides meaning. Relevance creates urgency. When those two are absent, people rarely rebel. They do not storm out. They stay, but they disconnect in small, quiet ways. And disconnection is the slowest way to lose people. When those two are present, something shifts. A check-in becomes trust. Feedback leads to growth. A vision creates energy. So here’s the reminder — whether you’re writing, designing, selling, or leading: Anchor in context. Deliver relevance. Only then will your story not just be told, but heard, remembered, and acted upon. Because without those two, you don’t have a story. You have a monologue. And on a Monday morning, that’s the real test: Are you speaking into the room’s reality — or into your own echo? #careershifts #context #relevance

Do you want to emphasize an important word in your sentence? Place that word at the end of the sentence. To do this, you may need to re-work your sentence structure. You may need to break up a long sentence into two (or three) shorter sentences. The effort will be worthwhile. Here’s an example: ▪️ The court must reject these arguments because the Highlands Act does not establish a deadline for the establishment of the TDR program. ✳️ These arguments must be rejected. The Highlands Act does not establish a deadline for the establishment of the TDR program. The second version is more effective for two reasons: ➖ It is easier to read broken into two sentences. ➖ The strong word “rejected” is highlighted by its placement at the end of the first sentence. But placing an important word at the end of the sentence is not always the best move. Look at this example: ▪️ Because XXX’s Supplement is an appropriate supplemental disclosure under Rule 26(e), and because Defendants submitted it well before the deadline for pretrial disclosures and within the expert discovery period, the Supplement is timely. Yes, “timely” is an important word here, and it’s emphasized by its position at the end of the sentence. But the sentence—with two dependent “because” clauses—is unwieldy. The reader needs to keep both dependent clauses in mind before getting to the final “the Supplement is timely”. The writer is making the reader work too hard. I’d re-work the sentence, flipping “the Supplement is timely” to the beginning. Here’s one way to do that: ✳️ The Supplement is timely for two reasons: (1) It is an appropriate supplemental disclosure under Rule 26(e), and (2) Defendants submitted it well before the deadline for pretrial disclosures and within the expert discovery period. Or perhaps use a vertical list for the two reasons. ✳️ The Supplement is timely for two reasons:  (1) It is an appropriate supplemental disclosure under Rule 26(e), and  (2) Defendants submitted it well before the deadline for pretrial disclosures and within the expert discovery period. Or, another possibility: Use a colon after “timely” to introduce the two reasons. ✳️ The Supplement is timely: It is an appropriate supplemental disclosure under Rule 26(e), and Defendants submitted it well before the deadline for pretrial disclosures and within the expert discovery period. The end of a sentence can be a powerful position. Please share an example of a sentence that is effective because the important word is placed at the end. Add whatever context is needed.

Over 80% of users skim, so when a PDP tries to say everything at once, it ends up saying nothing. A cluttered PDP gets more friction than function. Overwhelming users, leading to: - less time spent on page - missing value cues - fewer checkouts A well structured PDP doesn’t overwhelm, rather presents the information in a clear and digestible manner. Encouraging them to take action. In this post, I’ve broken down 12 changes I made to make the PDP easier to read and more focused on what actually helps users purchase. 1. Highlight customer satisfaction upfront. Show how many customers have purchased in the announcement bar. This builds immediate social proof that stays on all your pages. 2. Add benefit-focused badges above the product name. These help shoppers understand what key problems the product solves without needing to read through paragraphs. 3. Keep the title clear, and use a short subtitle to summarise the product and its core benefit. This helps users get both the “what” and the “why” at a glance. 4. Show the number of reviews beside the rating. It adds transparency and makes the rating feel more trustworthy, especially for first-time visitors. 5. Clarify price and pack size early. It saves users from searching for basic details which keeps attention focused on the purchase. 6. Use a context-rich main image. Featuring the product in its real-world use makes it easier to understand what’s being sold and how it fits into everyday life. 7. Expand image thumbnails beyond angles. Include images that show packaging and portion size to help customers evaluate fit and quality. 8. Add 2–3 bullet points above the fold. These help break down the product’s key benefits clearly, making it easier for skimmers to understand what makes it different. 9. Reinforce trust near the Add to Cart section. This is where buying hesitation happens so highlight things like delivery speed, return policies, or support to reduce friction. 10. Use icon-based highlights instead of long descriptions. Visual markers help users absorb information faster and keep the layout clean and scannable. 11. Break down product details visually. Showing ingredient percentages or content breakdowns in a simplified format helps make complex info more digestible. 12. Use accordions (not horizontal tabs). This allows users to expand only what they need, keeping the page organized and improving mobile usability. 13. Bring related variants closer to the decision zone. Show similar options earlier to help customers switch easily without needing to scroll to the bottom. Other UI/UX changes I did – Reduced text density to improve readability – Used consistent icons to simplify scanning – Added color cues for visual balance Found this useful? Let me know in the comments. PS: This checklist helps PDPs be clear and easy to follow without cramming in too much at once. This in turn will help the users make informed decisions that drive action. 

Your RAG pipeline is only as good as what it retrieves. And that’s exactly where most RAG chatbots quietly fail. You’re in a GenAI discussion, and someone asks: “Why does traditional RAG sometimes give confident but wrong answers?” RAG (Retrieval-Augmented Generation) assumes that the retrieved context is relevant and sufficient. But in reality, retrieval can be noisy, incomplete, or just plain wrong. And once bad context enters the pipeline, the LLM doesn’t question it. It just builds on top of it. That’s where Corrective RAG (CRAG) changes the game. What goes wrong in traditional RAG? 📍Retrieval returns low-quality or irrelevant documents 📍No mechanism to validate context before generation 📍LLM blindly trusts retrieved chunks Result → hallucinations with high confidence What CRAG does differently👇 CRAG introduces a correction layer between retrieval and generation. Instead of assuming retrieval is correct, it asks: 👉 “Is this context actually useful?” It does this through: 1. Retrieval Evaluation A lightweight evaluator (often a smaller model) scores the quality of retrieved documents. 2. Conditional Flow If retrieval is good → proceed as usual If retrieval is bad → trigger corrective actions 3. Corrective Actions Re-retrieve using refined queries Perform web search or external lookup Filter out noisy chunks Decompose the query for better context Traditional RAG is retrieve → generate CRAG is retrieve → evaluate → correct → generate #ai #rag #chatbot #retrieval #vectorsearch #aisystems #aiengineering Follow Sneha Vijaykumar for more...😊

Excited to share LinkedIn's innovative approach to evaluating semantic search quality! As part of the Search AI team, we've developed a groundbreaking evaluation pipeline that revolutionizes how we measure search relevance. >> Key Innovation: On-Topic Rate (OTR) This novel metric measures the semantic match between queries and search results, going beyond simple keyword matching. The system evaluates whether content is truly relevant to the query's intent, not just matching surface-level terms. >> Technical Implementation Details Query Set Construction • Golden Set: Contains curated top queries and complex topical queries • Open Set: Includes trending queries and random production queries for diversity Evaluation Pipeline Architecture 1. Query Processing: - Retrieves top 10 documents per query - Extracts post text and article information - Processes both primary content and reshared materials 2. GAI Integration: - Leverages GPT-3.5 with specialized prompts - Produces three key outputs:  - Binary relevance decision  - Relevance score (0-1 range)  - Decision reasoning Quality Assurance • Validation achieved 94.5% accuracy on a test set of 600 query-post pairs • Human evaluation showed 81.72% consistency with expert annotators >> Business Impact This system now serves as LinkedIn's benchmark for content search experiments, enabling: • Weekly performance monitoring • Rapid offline testing of new ML models • Systematic identification of improvement opportunities What are your thoughts on semantic search evaluation? 

🔓 Make Your CV a Magnet for Hiring Managers In today's competitive job market, your CV needs to stand out from the crowd. This can also be said about your LinkedIn profile. 🤔 But how? It's all about strategic layout and presentation. Let's unlock the secrets to impress hiring managers and land that dream interview! ✨ ✅ 1. Clarity is King: Structure: Use clear headings, bullet points, and white space for easy reading. Think scannable, not crammed. Conciseness: Keep it relevant and highlight achievements, not a novel. 3-4 pages max for experienced professionals. Shorter the better! Formatting Consistency: Stick to a professional font (e.g., Arial, Times New Roman) and font size (e.g., 10-12 pt). ✅ 2. Content is Key: Tailor It!: Adapt your CV/profile to each job description, emphasising relevant skills and experience. Quantify Your Impact: Use numbers and metrics to showcase achievements (e.g., "Increased sales by 20%"). Action Verbs: Highlight strong action verbs like "spearheaded," "achieved," or "implemented" to demonstrate initiative. ✅ 3. Presentation Polishes: Proofread, Proofread, Proofread!: Typos and grammatical errors are dealbreakers. Get a second pair of eyes on it. Invest in Visual Appeal: Use professional formatting tools or templates to create a sleek and modern look. PDF Power: Save your CV as a PDF to ensure consistent formatting across devices. 💥 Remember: Your LinkedIn profile and CV represent your personal brand story. Make it compelling, concise, and visually appealing to grab attention and secure that interview! 💡 What are your best CV tips? Share in the comments below #jobsearch #career #CV #resume #linkedin #hiring #tips #advice #interview

Many companies have started experimenting with simple RAG systems, probably as their first use case, to test the effectiveness of generative AI in extracting knowledge from unstructured data like PDFs, text files, and PowerPoint files. If you've used basic RAG architectures with tools like LlamaIndex or LangChain, you might have already encountered three key problems: 𝟭. 𝗜𝗻𝗮𝗱𝗲𝗾𝘂𝗮𝘁𝗲 𝗘𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 𝗠𝗲𝘁𝗿𝗶𝗰𝘀: Existing metrics fail to catch subtle errors like unsupported claims or hallucinations, making it hard to accurately assess and enhance system performance. 𝟮. 𝗗𝗶𝗳𝗳𝗶𝗰𝘂𝗹𝘁𝘆 𝗛𝗮𝗻𝗱𝗹𝗶𝗻𝗴 𝗖𝗼𝗺𝗽𝗹𝗲𝘅 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀: Standard RAG methods often struggle to find and combine information from multiple sources effectively, leading to slower responses and less relevant results. 𝟯. 𝗦𝘁𝗿𝘂𝗴𝗴𝗹𝗶𝗻𝗴 𝘁𝗼 𝗨𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗮𝗻𝗱 𝗖𝗼𝗻𝗻𝗲𝗰𝘁𝗶𝗼𝗻𝘀: Basic RAG approaches often miss the deeper relationships between information pieces, resulting in incomplete or inaccurate answers that don't fully meet user needs. In this post I will introduce three useful papers to address these gaps: 𝟭. 𝗥𝗔𝗚𝗖𝗵𝗲𝗸𝗲𝗿: introduces a new framework for evaluating RAG systems with a focus on fine-grained, claim-level metrics. It proposes a comprehensive set of metrics: claim-level precision, recall, and F1 score to measure the correctness and completeness of responses; claim recall and context precision to evaluate the effectiveness of the retriever; and faithfulness, noise sensitivity, hallucination rate, self-knowledge reliance, and context utilization to diagnose the generator's performance. Consider using these metrics to help identify errors, enhance accuracy, and reduce hallucinations in generated outputs. 𝟮. 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁𝗥𝗔𝗚: It uses a labeler and filter mechanism to identify and retain only the most relevant parts of retrieved information, reducing the need for repeated large language model calls. This iterative approach refines search queries efficiently, lowering latency and costs while maintaining high accuracy for complex, multi-hop questions. 𝟯. 𝗚𝗿𝗮𝗽𝗵𝗥𝗔𝗚: By leveraging structured data from knowledge graphs, GraphRAG methods enhance the retrieval process, capturing complex relationships and dependencies between entities that traditional text-based retrieval methods often miss. This approach enables the generation of more precise and context-aware content, making it particularly valuable for applications in domains that require a deep understanding of interconnected data, such as scientific research, legal documentation, and complex question answering. For example, in tasks such as query-focused summarization, GraphRAG demonstrates substantial gains by effectively leveraging graph structures to capture local and global relationships within documents. It's encouraging to see how quickly gaps are identified and improvements are made in the GenAI world.

The most underestimated part of building LLM applications? Evaluation. Evaluation can take up to 80% of your development time (because it’s HARD) Most people obsess over prompts. They tweak models. Tune embeddings. But when it’s time to test whether the whole system actually works? That’s where it breaks. Especially in agentic RAG systems - where you’re orchestrating retrieval, reasoning, memory, tools, and APIs into one seamless flow. Implementation might take a week. Evaluation takes longer. (And it’s what makes or breaks the product.) Let’s clear up a common confusion: 𝗟𝗟𝗠 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 ≠ 𝗥𝗔𝗚 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻. LLM eval tests reasoning in isolation - useful, but incomplete. In production, your model isn’t reasoning in a vacuum. It’s pulling context from a vector DB, reacting to user input, and shaped by memory + tools. That’s why RAG evaluation takes a system-level view. It asks: Did this app respond correctly, given the user input and the retrieved context? Here’s how to break it down: 𝗦𝘁𝗲𝗽 𝟭: 𝗘𝘃𝗮𝗹𝘂𝗮𝘁𝗲 𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹. → Are the retrieved docs relevant? Ranked correctly? → Use LLM judges to compute context precision and recall → If ranking matters, compute NDCG, MRR metrics → Visualize embeddings (e.g. UMAP) 𝗦𝘁𝗲𝗽 𝟮: 𝗘𝘃𝗮𝗹𝘂𝗮𝘁𝗲 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻. → Did the LLM ground its answer in the right info? → Use heuristics, LLM-as-a-judge, and contextual scoring. In practice, treat your app as a black box and log: - User query - Retrieved context - Model output - (Optional) Expected output This lets you debug the whole system, not just the model. 𝘏𝘰𝘸 𝘮𝘢𝘯𝘺 𝘴𝘢𝘮𝘱𝘭𝘦𝘴 𝘢𝘳𝘦 𝘦𝘯𝘰𝘶𝘨𝘩? 5–10? Too few. 30–50? Good start. 400+? Now you’re capturing real patterns and edge cases. Still, start with how many samples you have available, and keep expanding your evaluation split. It’s better to have an imperfect evaluation layer than nothing. Also track latency, cost, throughput, and business metrics (like conversion or retention). Some battle-tested tools: → RAGAS (retrieval-grounding alignment) → ARES (factual grounding) → Opik by Comet (end-to-end open-source eval + monitoring) → Langsmith, Langfuse, Phoenix (observability + tracing) TL;DR: Agentic systems are complex. Success = making evaluation part of your design from Day 0. We unpack this in full in Lesson 5 of the PhiloAgents course. 🔗 Check it out here: https://lnkd.in/dA465E_J

On taking time to do a good lit review & craft a strong paper (or you can't take shortcuts). As a senior editor, I read dozens of reviews a year. Often, a reviewer complains that the authors have failed to cite papers or miscited papers. Sometimes, they use that lack of due diligence to reject a paper. I wrote about the problem of rejection bc of missing papers here - * took a strong position against doing so: https://lnkd.in/eAHSZKW5 However, I want to offer a clarification for early career authors. Calling for reviewers to educate & help authors, does not mean that you should submit papers that are under-researched or quickly assembled. It means that errors of omission occur even in well-researched & thoughtfully assembled papers. It is up to you, the authors, to do your very best job to stay up to date on the literature, write reasonable arguments, & submit work using current methods. This is easy to write - but hard to do - so how to ensure you've done a good job? First, check for updated references before submission. Papers take months to write. It's natural a few relevant citations may pop up once you've completed the literature review. I like to check for new references to key citations in my paper. I do so on Googlescholar.com. If I find relevant references, I add them to the paper. Second, edit once, edit twice, edit thrice. Early career authors often submit papers with small grammatical errors. I've asked a few of them why? They say that grammar should not get in the way of seeing the ideas. Clearly, they have not met my OCD friend Reviewer Two. Grammar matters. Third, check author names. Grammar & spelling checkers won't catch author name errors. In fact, they often amend them to be incorrect. If your checker recommends an alternative spelling, check the reference. Nothing makes a reviewer crabbier than seeing their name misspelled - it's happened to me! Fourth, really take time to know your topic. In the high-volume, high-pressure world that we work in, many authors are submitting papers on diverse topics. That they have a superficial understanding of a topic shows up in the literature review. The early stages of your work on a new topic should involve a few weeks, if not months of reading. Note, bc of this high overhead, I suggest people write synergistic studies or sets of papers on related topics or questions. Finally, have a knowledgeable peer read the paper before submission. The issue isn't just having the right citations, as Bryan Gaensler points out, it takes time to understand the nuances of a literature. Have a second set of eyes read the paper to make sure that not only the right citations are there but that they are used in the right spirit. If you take time to update references, carefully edit the paper, & solicit a knowledgeable peer review, you will have much more luck in securing a revision at a good journal.