Songya Pan

(2/3) Why agents fail. How to fix them. 🧠 Reasoning errors are the main way agents fail on the DA-Code benchmark ⚡ Injecting a strong evaluator for inference-time improvement led to +30% higher task success Full blog post in comments.

24

1 Comment

Another good one from Karpathy’s an 1-hour LLM primer (I missed to watch earlier) covers: • How LLMs are trained (and why the weights matter) • Fine-tuning into helpful assistants • Tool use & multimodality in action • Scaling laws driving progress • Security risks: jailbreaks, prompt injection, data poisoning https://lnkd.in/ev38636z

4

Excited to share that our team’s paper on aspect-guided review summarization has been accepted to EMNLP 2025’s industry track! 📈🧠🤓 In this paper, we describe a live system at Wayfair that leverages LLMs alongside classical NLP techniques to summarize tens of millions of customer reviews, making it easier for shoppers to discover what matters most to them on site. To further support the research community, we’re also open-sourcing a dataset of ~12M anonymized reviews and their extracted aspects which we hope accelerates progress in aspect extraction, sentiment analysis, and summarization tasks. This milestone reflects the hard work of many folks across Wayfair, including our co-authors Ilya Boytsov, Misha Balyasin, Joe Walt, Caitlin Eusden, Marie-Claire Rochat, and Margaret Pierson. https://lnkd.in/ezJMi3g7 #NLP #EMNLP #EMNLP2025 #LLMs #Summarization #ABSA #OpenSource #Data #Dataset #ecommerce #wayfair #machinelearning #ai #publication

74

3 Comments

I can see the future of autonomous taxis as they would certainly help with peak times demand and other use cases, but I can also see how there will be many challenges of making it really work at scale. Here is one, a very ''analogue'' problem, which I am sure leaves Waymo engineers wake at night. Local fleet management operations are crucial, meaning, in every city where AV taxis are launching, the underlying infrastructure forms the backbone and ensures the fleet runs smoothly.

235

74 Comments

Third piece in the AAD series is up. This one is about acceptance criteria (AC) and why treating them as a formal contract changes what you can diagnose when autonomous development breaks.  AC serve as the bridge between the spec and the environment in which the spec is compiled into a product. This is implicit in the first article, but after marinating on this for a few weeks, I felt like I should make it explicit.                                                                                                    The short version: most teams write AC the way you'd write them for a human. "Handle edge cases gracefully." "Performance should be acceptable under normal load." In AAD, those are  specification failures. An agent will do something when it reads that. It might even do something reasonable. But you've handed control over what "reasonable" looks like to the LLM, and you'll only find out whether you agree after you've burned the iteration cycle. In my experience, the first pass is ok, but it can quickly diverge and it takes a lot of time to get back to where you intended to go in the first place. One thing I keep coming back to is the diagnostic value of keeping the spec and the AC separate. My currently line of thinking leads me to this framing: The spec describes intent. The AC defines done. You need both. The Spec is the what and the why. The AC are the falsifiable tests that the intention was met. Keeping them separate creates a framework so they each get the attention they deserve as part of this new world. Building on the notion of falsifiability, there's also a section on the right mindset for writing AC. Just like with the best scientist/experimentalist, you should be writing AC from a lens that lets you confirm intention was not met, not that intention was met. It's a subtle but meaningful nuance. If you assume the agent is going to get it correct, you're biased to see the final product and assume it's correct. When you train yourself to focus on the ways the agent might get it wrong, you open up a whole new world on writing great AC. I think this applies to traditional development as well as AAD. Claude made a fancy doc for this one! But also, the full read is in the comments.

15

1 Comment

Front-loaded prompts, Iterative prompts and now Reverse Prompts I guess most of the scenarios, we interact with LLMs in two common ways: Front-loaded prompting – we carefully craft a long, detailed context and hope the model delivers the exact output and Iterative prompting – we start with a simple question and keep refining, nudging the model until it gets closer to what we want. Both work, but sometimes Reverse Prompting is more powerful. (haven't came across any industry standard name for this - please comment if any) In reverse prompting, instead of me asking all the questions, I asked the model to ask me the right questions first. Think of it like visiting a doctor. You don’t walk in and give a 3-page monologue of your symptoms. The doctor asks: Do you have a headache? Is the pain sharp or dull? Does it get worse at night? Only an expert knows the right sequence of questions that lead to the right diagnosis. I simulated this with Chat GPT. I set its role as an orthopedic specialist and said you have to ask question like a doctor before prescribing. Then I simply said - “I have pain in my lower back.” Instead of rushing into advice, it asked me a sequence of questions about lifestyle, posture, history, even suggested I provide X-ray results. After a long reverse Q&A, it finally summarized possible causes and logical next steps. Honestly, it felt close to a real consultation (minus the waiting room), though in reality I am not replacing real doctor yet! What struck me is how powerful reverse prompting can be when we don’t have the expertise to ask the right questions ourselves. The AI guides the conversation, fills in the blanks, and leads us toward clarity. Maybe this is the future of working with AI: not just us prompting it, but it prompting us back — like a partner who knows what to ask, even when we don’t.

50

2 Comments

🚨 Two AI Legends Just Said the Same Thing: LLMs Aren't the Path to True Intelligence Richard Sutton (Father of Reinforcement Learning, 2024 Turing Award Winner) and Yann LeCun (Meta's Chief AI Scientist) are both moving beyond LLMs. Here's why this matters for everyone: 🧵 The Problem They Both See 🤔 Current AI chatbots (like ChatGPT) are basically sophisticated copy-cats. They: Mimic what humans would say ✍️ Don't actually understand consequences Can't truly learn from mistakes Have no real goals beyond predicting text Think of it like this: A parrot 🦜 can repeat "the sky is blue" perfectly, but it doesn't understand what sky, blue, or "is" means. What Real Intelligence Needs 🧠 Both experts say we need AI that: 1️⃣ Learns like a child does 👶 Not by copying, but by trying things and seeing what happens 2️⃣ Understands the physical world 🌍 Can predict: "If I push this cup, it will fall" 3️⃣ Has actual goals 🎯 Not just "predict the next word" but "achieve this specific outcome" 4️⃣ Plans ahead ♟️ Like chess players thinking multiple moves ahead, not just reacting The Future They Envision 🔮 Instead of bigger chatbots, imagine AI that: Learns continuously from experience 📈 Updates its understanding in real-time Can actually reason about cause and effect Builds mental models of how things work Why This Matters to You 💡 We're potentially at a turning point. The current AI boom might hit a wall 🧱, and the next breakthrough will come from completely different approaches. For businesses: Don't just bet on scaling current AI - watch for fundamentally new architectures For careers: Understanding real-world interaction and planning might become more valuable than prompt engineering The Bottom Line 📝 Two of AI's most influential thinkers are saying the same thing: We've been building really impressive calculators 🖩 when we should be building systems that actually understand the world. The next phase of AI won't just be ChatGPT-5 or GPT-6. It might be something entirely different. 🚀 What do you think - are we heading toward a major shift in how we approach AI? source: https://lnkd.in/gsHZvNbE https://lnkd.in/gchsKwsE #AI #GenAI #AGI #LLM #

41

3 Comments