Top LLM Papers of the Week (October Week 4, 2024)

This week's newsletter is sponsored by Dynamiq (open-source Gen AI framework)

[1] O1 LLM Replication Journey

This paper presents the details of the project to replicate the capabilities of OpenAI’s O1 LLM. The paper emphasizes the limitations of? “Short Cut Learning” and introduces a new paradigm called “Journey Learning”. Journey learning enables to develop more robust AI systems by allowing them to explore the entire problem-solving process, including trial and error, reflection, and backtracking.??[Tweet] and [Paper]

[2] GPT-4o System Card

This paper presents GPT-4o details. GPT-4o is an autoregressive omni model, which accepts as input any combination of text, audio, image, and video and generates any combination of text, audio, and image output. It’s trained end-to-end across text, vision, and audio, meaning that all inputs and outputs are processed by the same neural network. GPT-4o matches GPT-4 Turbo performance while also being much faster and 50% cheaper in the API. [Tweet] and [Paper]

Dynamiq is an all-in-one open-source Gen AI framework. Dynamiq can be used to streamline the development of LLM-powered applications.

Use cases of Dynamiq framework

- Create LLM Flows

- Create ReAct Agent

- Multi-agent orchestration

- Create RAG Applications

- Create Chatbot with Memory

"Top LLM Papers of the Week" newsletter is read by over 20k+ AI Researchers, Engineers and Developers. If you would like to promote with us, contact Kalyan KS

[3] AutoRAG

This paper introduces the AutoRAG framework, which automatically identifies suitable RAG modules for a given dataset. AutoRAG explores and approximates the optimal combination of RAG modules for the dataset.? AutoRAG is available as a Python library.? [Tweet] and [Paper]

[4] Parameter-Efficient Fine-Tuning in LLMs (Survey)

This paper presents a comprehensive survey of parameter efficient fine-tuning methods for LLMs. Specifically, the paper introduces the preliminary knowledge of PEFT, the core ideas and principles of various PEFT algorithms, the applications of PEFT, and potential future research directions. [Tweet] and [Paper]

[5] Small Language Models (Survey)

Small Language Models (SLM) are gaining attraction because of their ability and efficiency to perform various tasks with minimal computational resources. This paper presents a comprehensive survey on SLMs, focusing on their architectures, training techniques, and model compression techniques.? The paper also introduces a novel taxonomy for categorizing the methods used to optimize SLMs, including model compression, pruning, and quantization techniques.?[Tweet] and [Paper]

[6] ChunkRAG

This paper introduces ChunkRAG, a novel chunk filtering method for RAG systems. Existing RAG methods fail to effectively filter irrelevant chunks which can results in? inaccurate responses.? ChunkRAG? employs semantic chunking to divide documents into coherent sections and utilizes LLM-based relevance scoring to assess each chunk’s alignment with the user’s query. Experiments demonstrated that ChunkRAG outperforms existing RAG models.? [Tweet] and [Paper]

[7] Benchmarking LLM Compression Methods

Many model compression techniques are proposed to increase the efficiency of LLMs.? This paper presents LLMCBench, a benchmark with an in-depth analysis for LLM compression algorithms. LLMCBench can contribute insightful suggestions for LLM compression algorithm design and serve as a foundation for future research. [Tweet] and [Paper]

[8] Specialized MOEs For Financial Analysis

This paper presents FinTeamExperts, a role-specialized LLM framework structured as a Mixture of Experts (MOEs) for financial analysis. The framework simulates a collaborative team setting by training each model to specialize in distinct roles: Macro Analysts, Micro analysts, and Quantitative Analysts.? This role-specific specialization enhances the model's ability to integrate their domain-specific expertise. [Tweet] and [Paper]

[9] Mistral-based Medical LLM

This paper introduces? BioMistral-NLU an LLM for medical domain. BioMistral-NLU is developed by fine-tuning BioMistral over a curated instruction dataset. Experiments results show that BioMistral-NLU outperforms the original BioMistral, as well as the proprietary LLMs - ChatGPT and GPT-4. [Tweet] and [Paper]

[10] Multi-Agent Framework for Data Science Competitions

This paper presents AutoKaggle, a powerful and user-centric framework that assists data scientists in completing daily data pipelines through a collaborative multi-agent system. AutoKaggle implements an iterative development process that combines code execution, debugging, and comprehensive unit testing to ensure code correctness and logic consistency. AutoKaggle achieves a validation submission rate of 0.85 and a comprehensive score of 0.82.? [Tweet] and [Paper]

Do subscribe to the newsletter so that you won't miss interesting updates related to Generative AI, LLMs and RAG.

Kalyan KS, Research Scientist(NLP) at Akmmus AI Labs