Zak Jost

This is really awesome. I used notebookLM to summarise a recent paper that shows how o1 like model can be reproduced. After providing the pdf from Arxiv, I noticed that the generated podcast was almost 49 minutes long. Very different from the earlier 10-12 minute version. And another impressive feature was the interactive mode where you can interject the two hosts and ask a question which they very graciously respond to. I wanted to share the audio summary but YT didn’t allow me to upload the WAV file. YT expects a video mp4 file. So I found free videos from Pexels and stitched them together with ffmpeg. The videos are meditative acquarium shots of fish swimming around. And with the audio discussing reproducing o1 as an engaging podcast episode, the video keeps the distraction to the minimum. The paper itself is a great read but if you have a 30-40 minute commute, or want to watch some cool fish swimming around and learn about o1, check out

12

🚨 New paper w/ interesting insights for fine-tuning and GPU poor folks 🚨 🚀 Exploring Fine-Tuning for LLMs: Full Fine-Tuning vs. #LoRA - Are They Truly Equivalent? 🤔 As LLMs grow, fine-tuning them for specific tasks becomes challenging due to high computational costs. Recently, Low-Rank Adaptation (LoRA) has become a popular alternative, promising similar performance with fewer parameters. But are LoRA-tuned models the same as fully fine-tuned ones? MIT’s latest study dives deep into this question, revealing some surprising differences! 🔍 📝Key Takeaways: 1. Spectral Differences 🧬: LoRA introduces “intruder dimensions” not found in fully fine-tuned models. These alter the model’s weight structure, suggesting that LoRA and full fine-tuning leverage different parts of the parameter space, even if they hit similar accuracy on specific tasks. 2. Generalization & Robustness 🌐: Models fine-tuned with LoRA may struggle with new or sequential tasks, forgetting previously learned knowledge more easily. Full fine-tuning, however, shows better retention and robustness in shifting environments. 3. Effective Rank & Optimal Performance ⚙️: LoRA’s rank matters! Mid-range ranks (like r=64) perform well on tasks and retain pre-trained knowledge, while lower or higher ranks can lead to greater forgetting. Smart parameter tuning (like setting α=2r) can help LoRA resemble full fine-tuning more closely. 4. Why It Matters 💡: LoRA offers incredible parameter efficiency, but its unique trade-offs might affect its use in scenarios needing high adaptability and long-term memory. With adjustments, LoRA can bridge some of these gaps, but full fine-tuning remains the go-to for complex or multi-tasking demands.

25

1 Comment

🔮RAG + KG: Powerful Combination for Translation Multi-AI Agents Large language models (LLMs) have shown great promise in machine translation, but they still struggle with contextually dependent terms, such as new or domain-specific words. This leads to inconsistencies and errors that are difficult to address. Existing solutions often depend on manual identification of such terms, which is impractical given the complexity and evolving nature of language. While RetrievalAugmented Generation (RAG) could provide some assistance, its application to translation is limited by issues such as hallucinations from information overload. =================== In this paper, researchers propose CRAT, a novel multi-agent translation framework that: 》 leverages RAG and causality enhanced self-reflection to address these challenges. This framework consists of several specialized agents: ✱ the Unknown Terms Identification agent detects unknown terms within the context, ✱ the Knowledge Graph (KG) Constructor agent extracts relevant internal knowledge about these terms and retrieves bilingual information from external sources, ✱ the Causality enhanced Judge agent validates the accuracy of the information, ✱ Translator agent incorporates the refined information into the final output. =================== This automated process allows for more precise and consistent handling of key terms during translation. Their results show that CRAT significantly improves translation accuracy, particularly in handling context-sensitive terms and emerging vocabulary. --- If your company is looking for an AI consultant to accelerate business growth, or if you are looking to build your career in AI, send me a message and I can help

4

A very useful checklist to follow when you have to put in production Transformer based models

5

"Prompt Engineering" by Lee Boonstra. This paper covers everything from basic prompting techniques to advanced strategies like Chain of Thought (CoT), Tree of Thoughts (ToT), and ReAct, along with practical tips for code-related prompting. Key sections include: - Configuring output with temperature, top-K, and top-P settings - Using zero-shot, one-shot, and few-shot prompting effectively - System, role, and contextual prompting techniques - Best practices, including experimenting with input formats and adapting prompts to model updates https://lnkd.in/gH5xrivh

13

LLMs have become more capable with better training and data. But they haven’t figured out how to “think” through problems at test-time. The latest research from Scale finds that simply scaling inference compute–meaning, giving models more time or attempts to solve a problem–is not effective because the attempts are not diverse enough from each other. 👉 Enter PlanSearch, a novel method for code generation that searches over high-level "plans" in natural language to encourage response diversity. PlanSearch enables the model to “think” through various strategies before generating code, making it more likely to solve the problem correctly. The Scale team tested PlanSearch on major coding benchmarks (HumanEval+, MBPP+, and LiveCodeBench) and found it consistently outperforms baselines, particularly in extended search scenarios. Overall performance improves by over 16% on LiveCodeBench from 60.6% to 77%. Here’s how it works: ✅ PlanSearch first generates high-level strategies, or "plans," in natural language before proceeding to code generation. ✅ These plans are then further broken down into structured observations and solution sketches, allowing for a wider exploration of possible solutions. This increases diversity, reducing the chance of the model recycling similar ideas. ✅ These plans are then combined before settling on the final idea and implementing the solution in code. Enabling LLMs to reason more deeply at inference time via search is one of the most exciting directions in AI right now. When PlanSearch is paired with filtering techniques—such as submitting only solutions that pass initial tests—we can get better results overall and achieve the top score of 77% with only 10 submission attempts. Big thanks to all collaborators on this paper including: Evan Wang, Hugh Zhang, Federico Cassano, Catherine Wu, Yunfeng Bai, William Song, Vaskar Nath, Ziwen H., Sean Hendryx, Summer Yue 👉 Read the full paper here: arxiv.org/abs/2409.03733

180

2 Comments

Amazing performance boost by DeepSpeed. FPx speed-ups on training and inference is always preferable to INTx quantization.

9

In context of debate if large context #LLM models will eventually negate the need for #RAG, an article from Databricks has some good insights - • Longer contexts boost RAG accuracy... to a point • Performance often peaks at 32k-64k tokens, then declines • Top models (GPT-4, Claude 3.5) maintain consistency at scale • Others show unique failure modes (e.g., copyright concerns, summarization instead of Q&A) • Optimal context size varies by model and task • Lack of long-context training data may explain some issues Key takeaway: Long context and RAG are synergistic, but LLMs need refinement to fully leverage extended contexts. https://lnkd.in/eehdKtCm

17

Notion vs Glean reminds me a lot of Tesla vs Waymo... Just like I think Tesla is going to solve the 'software' problem of autonomous vehicles before Waywmo solves the 'hardware' problem... I think that Notion is going to solve 'enterprise knowledge search' before Glean solves the problem of document and knowledge creation... Why? A couple reasons: 1) Just like Tesla, Notion has a huge advantage when it comes to the amount of underlying data and documents to train on... They already have 100M+ users, and likely 10x that in terms of the number of documents, files, etc. in team's workspaces. 2) It's very hard to change people's behaviour re: where/how they create 'work'. It's why a lot of companies have come and gone in that category... As exciting as Glean's AI writing assistants might seem, most people are just fine (or even prefer) a good old fashioned editor/writer. I could be way off here, but this will be an intersting one to follow. Thoughts? Where am I getting it wrong? #ai #enterprise #knowledge #management #notion #glean

3

1 Comment

“There is no wall” - Sam Altman Frontier Math - tackling extremely hard mathematical problems - achieves normal accuracy of about 2% with current models. However, O3 demonstrates a remarkable 25%, suggesting that it possesses domain-specific expertise. Terence Tao, a highly regarded mathematician, once stated that solving such problems would require deep domain expertise. The performance of O3 clearly indicates its capability in this regard. The speed at which the rest of the world is reacting to these advancements is incomparable to the progress seen in just the last three months. OpenAI released O1 only three months ago, and now O3 has already raised the bar. Consider the SWE benchmark: the best rating so far was 49%, but O3 has achieved an impressive 71%. 💡Only a few weeks ago, I found myself wondering whether any model could transcend the intelligence of luminaries in specific fields. Today, I think it is indeed possible. ❓Have we reached a state of AGI? While there are still elements to consider, it feels closer now than ever before. What are your thoughts? #AI #OpenAI #O3Model #AGI #TechInnovation

3

3 Comments

Origins of Planning explores, how biological brains plans, interferes complex calculations and how machines can tackle hard problems in the cloud of probability. https://lnkd.in/gT2T2wS2

8

Buckle up, folks. Act 1 at Machina Labs was just the opening riff. We started with Roboforming; not because it was easy, but because it was the hardest. Robots cal, force measurement, scanning, deflection compensation, data pipelines, visualization, process modeling, high-speed trajectory management, etc. we developed them all for Roboformig. Now that we learned the hardest, the rest is just remixing the features. Case in point: Robosanding. Added, tested, and running in under a week. The future? A system handling 10-15 operations autonomously isn’t sci-fi; it’s a pit stop away. Watch this space.

214

19 Comments

Sometimes model training is expensive. How do you balance between reprodcibility and efficiency? 1. When you are training traditional ML models that don't take long enough to train (few mins to few hours). ✅ Retrain your model using the training pipeline rather than re-using the model trained by notebooks. Even if it gives the same results. 2. When your model takes several hours or days to train ✅ Do experimentation on smaller dataset and do full run only using pipelines. This doesn't have to be the case for deployments in lower environment. It is fine to deploy notebook generated models during development but make sure you are able to reproduce the same results on the same data using the training pipeline before you move to production.

5

Thanks Werner Bogula for the really insightful study. I’m only just getting into model comparison, for now using HuggingChat and some tightly defined use cases. The difference between them is notable, interesting to see, and makes me think the future is probably going to be based around multi-LLM workflows. #genai #multillm

1

- SwiftKV is a novel model transformation and distillation procedure (by folks at Snowflake) designed to optimize large language model (LLM) inference. Very cool findings in the paper (as presented in the vLLM meetup on 11/13). - SingleInputKV: Prefills later layers' KV cache using earlier layer outputs, reducing computation on prompt tokens. The key point is that the first few layers of attention in LLMs are similar (nice find), allowing SwiftKV's SingleInputKV mechanism to skip computations in later layers during inference. By using a single earlier layer's output to compute the KV cache for subsequent layers, it significantly reduces both computation (by 50%) and memory usage (by 62.5%) without sacrificing model performance. - AcrossKV: Merges KV caches of neighboring layers to minimize memory usage and increase batch size for higher throughput. - For models like Llama-3.1-8B and 70B, SwiftKV reduces prefill compute by 50%, memory usage by 62.5%, and achieves up to 2× higher throughput and 60% lower time per token. https://lnkd.in/guhEBW5G

62