Shankar Sastry

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

🔍 Quantum AI Series: Modeling Sequences with Quantum States Picking up where I left off in my #QuantumAI series, let’s talk about Modeling Sequences with Quantum States, with insights from a fascinating paper on the topic. Classical probability distributions can be modeled using quantum states—particularly entangled states that retain information from subsystems. This approach allows quantum models to efficiently represent sequences without losing essential details. Tensor networks play a key role in optimizing the process, providing a compact, low-rank factorization of data. As I've shared previously, Singular Value Decomposition (SVD) is foundational to these models, as it helps represent the complex correlations present in sequential data. Matrix Product States (MPS) capture these rich interactions by using reduced densities to efficiently compress sequence representations while preserving their structure. This is the essence of keeping representation while building efficient lightweight models. In summary, quantum probability models allow us to go beyond classical limits in sequential tasks, contributing to information retention. Sharing the paper for a deeper dive if you'd like so: https://lnkd.in/d5aqXyyS #QuantumAI #TensorNetworks #SVD #QuantumComputing #AI

5

🔗 Simulate PyTorch Models with Simulink! 🚀 Engineers, unlock new possibilities by simulating PyTorch and other Python-based models directly in Simulink using co-execution blocks. This integration streamlines your workflow, enabling seamless model testing and validation. 🔍 What's Inside: >> Co-execution of Python models in Simulink << >> Enhanced simulation and validation processes << 👉 Dive into the details: https://spr.ly/6041o1cOX #DeepLearning #Simulink #Python #ModelSimulation #Engineering #MATLAB #MAthWorks #PyTorch

9

[research update] use a Transformer's multi-head attention activations to fetch RAG style, to augment multi faceted problem breakdown for LLMs in ."Multi-Head RAG: Solving Multi-Aspect Problems with LLMs": 1. Enhances LLMs by integrating document retrieval (RAG) into the LLM context. 2. Challenge: Existing RAG solutions struggle with queries needing multiple documents with varied contents due to distant embeddings. 3. Multi-Head RAG (MRAG): Introduces a novel approach using Transformer's multi-head attention activations to fetch diverse documents. 4. Activation Utilization: Different attention heads capture varied data aspects, improving complex query retrieval. 5. Improved Relevance: MRAG shows up to 20% improvement in relevance over standard RAG baselines. 6. Evaluation Methodology: Provides metrics, synthetic datasets, and real-world use cases to demonstrate effectiveness. 7. Seamless Integration: MRAG can be integrated with existing RAG frameworks and benchmarking tools like RAGAS. 8. Multi-Aspect Problem Solving: Designed specifically to handle multi-aspect queries more effectively. 9. Data Stores Compatibility: Compatible with various data store classes. 10. Potential Applications: Benefits scenarios requiring diverse information retrieval, enhancing LLM responses. Experiment The paper introduces MRAG and evaluates it using synthetic datasets and real-world data. It compares MRAG's performance against standard RAG models. The experiment focuses on the model's ability to retrieve relevant documents across various query aspects, demonstrating up to a 20% improvement in relevance over baselines. Use Case MRAG is tested on real-world scenarios like customer support, where queries often require information from multiple documents with diverse contents. The model's capability to retrieve and integrate multi-aspect data enhances the quality of responses.

81

5 Comments

This article described AI as another "dot-com bubble" that is ready to bust anytime soon as soon as the "right time" comes and likened to a keg of gun powder waiting to explode at the next ignition opportunity. Hmmmmm ...... While some commercial entities might have portrayed AI as a magic wand that MUST work everywhere and take over ALL workflows, the reality is that it is a tool that has the capability to enhance existing workflows, create new opportunities to improve workflows, and support domain experts to do more and perform better at what they know to do best. Since AI, and especially the computational aspect of it, ML, relies on real data and the verifiable relationship between input features and the target properties, there is no opportunity to scam or spook anyone. The algorithms are well established in mathematics and statistics. There is a standard practice procedure (SPP) to machine learning applications that, if followed, would help to ensure the success in a modeling task or explain why it failed. It's not a magic wand. It's backed up by the physics, chemistry, and biology embedded in operational data. Success stories abound everywhere and, as a researcher, the reasons for the failed cases can be identified and mitigated. The article is here for those interested in more details: https://lnkd.in/dK9HYwn7 #artificialintelligence #ai #machinelearning #ml #dotcombubble

10

Modeling in the sciences is in fact almost always generative modeling. What a BOLD statement to start a paper with here --> 1.1 Motivation One major division in machine learning is generative versus discrimi- native modeling. While in discriminative modeling one aims to learn a predictor given the observations, in generative modeling one aims to solve the more general problem of learning a joint distribution over all the variables. A generative model simulates how the data is generated in the real world. “Modeling” is understood in almost every science as unveiling this generating process by hypothesizing theories and testing these theories through observations. For instance, when meteorologists model the weather they use highly complex partial differential equations to express the underlying physics of the weather. Or when an astronomer models the formation of galaxies s/he encodes in his/her equations of motion the physical laws under which stellar bodies interact. The same is true for biologists, chemists, economists and so on. Modeling in the sciences is in fact almost always generative modeling

1

1 Comment

We just released a blog on whether the use of multimodal feature transfer can detect deception (phenomena like sarcasm, irony, and condescension) in conversation. The short answer... Yes. This particular write-up profiles a subset of our work at Symbl.ai on multimodal large language models, and is research led by one of our summer interns, Benjamin R. (from Prof. Larry Heck's AVA Lab at Georgia Institute of Technology). We have a paper currently under submission on this work, and we can't wait to share the full details with all of you soon. Blog Link: https://lnkd.in/g6m_wFDi #multimodal #featuretransfer #LLMs #deception #conversation #sarcasm #irony #language #AI #ML #NLP #LargeLanguageModels #SummerInternship #ResearchPapers #ACL #Llama2 #SymblDotAI

16

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

Unlock the full potential of your Large Language Models (LLMs) with Intel® Extension for PyTorch (IPEX) and the Intel® LLM Library for PyTorch (IPEX-LLM). Download this whitepaper to explore the optimization of LLMs, including their performance, resource utilization, and response times in real-world applications. Link - https://intel.ly/3BBJ4ey

2

Brain Teaser #11: What mathematical arguments can explain how **Layer Normalization** enhances training stability in large language models (LLMs) and Transformer architectures, particularly in mitigating issues like vanishing and exploding gradients? Answer: Training destabilization in neural networks, particularly due to gradient explosion or vanishing, typically occurs because of the high condition numbers of weight matrices during backpropagation. These large condition numbers lead to unstable gradient scaling, either amplifying gradients excessively or diminishing them, hindering effective learning. How Layer Normalization overcomes and brings stability in training is briefly discussed here: https://lnkd.in/gmWm42tQ #MachineLearning #DeepLearning #Transformers #LargeLanguageModels #LayerNormalization #GradientStability #AIResearch #DataScience #NeuralNetworks #AIInsights

16

5 Comments

Looks like fine tuning out performs RAG and Large context LLMs for data accuracy with longer contexts Paper: https://lnkd.in/gj9-3jsK Summary of "BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystack" - Introduction to BABILong Benchmark: - Designed to test LLMs' reasoning abilities over extremely long documents. - Comprises 20 diverse reasoning tasks, such as fact chaining, induction, deduction, counting, and handling lists/sets. - Uses natural text from the PG19 corpus, extendable to document lengths up to 1 million tokens. - Current LLM Performance: - Popular LLMs utilize only 10-20% of the context effectively. - Performance declines with increased reasoning complexity. - Retrieval-Augmented Generation (RAG) methods achieve 60% accuracy on single-fact QA, independent of context length. - Top Performing Models: - Recurrent Memory Transformers (RMT) show the best performance, processing up to 11 million tokens. - Small models like Mamba (130M) and fine-tuned RMT (137M) achieve high accuracy on long-context tasks. - Evaluation Findings: - Current benchmarks (e.g., Longbench, L-Eval) are insufficient for LLMs with capabilities beyond 40,000 tokens. - BABILong tests models' efficiency in distinguishing relevant facts from irrelevant text over long contexts. - Popular LLMs fail to maintain high performance as context length and complexity increase. - Fine-Tuning Insights: - Fine-tuning improves model performance significantly. - Mamba and RMT demonstrate successful QA with long contexts, outperforming RAG models. - RMT's memory mechanism allows efficient processing of long sequences, with marginal quality degradation up to 11 million tokens. - Benchmark Significance: - BABILong provides a rigorous evaluation framework for LLMs' long-context reasoning abilities. - Highlights the limitations of current LLMs and the need for improved context processing mechanisms. - Key Contributions: - Introduction of BABILong, a scalable generative multi-task benchmark. - Evaluation of over 20 recent LLMs on various context lengths and tasks. - Demonstration of the limitations of current models in long-context utilization. - Setting a new record for sequence size processed by a single model with RMT.

4

5 Comments

Google quietly announced yesterday, with a tweet/X message from Sundar, the release of their new quantum computing chip/system...Willow. Supposedly it is able to handle a much larger number of qubits without errors. They got it to solve a computation that would require a standard supercomputer over 10 septillions (10,000,000,000,000,000,000,000,000) years. The quantum computer solved it in less than 300 seconds!!..... Let that sink in for an extra second... Marry that quantum computer up with an advanced AI model and there is nothing impossible anymore.

54

6 Comments

A recent breakthrough titled "Matrix Multiplication-Free LLMs" demonstrates a huge advancement in the area of Large Language Models (LLMs) by reducing computational costs. The authors have eliminated MatMul operations from LLMs, claiming to 10 times reduction in memory usage and a 25.6% increase in training speed, all while maintaining strong performance at billion-parameter scales. Paper link: https://lnkd.in/ggph8qXc #AI #machinelearning #deeplearning #LLMs

15