Papers Read on AI

The podcast explores the development and limitations of intelligent personal assistants, introduces the concept of Personal LLM Agents as a major software paradigm for personal computing devices, discusses their architecture, capability, efficiency, and security. It also covers the capabilities and challenges of lightweight language modeling agents, explores task-oriented dialogue systems and mobile task automation using LLMs, addresses the challenges of implementing practical personal LLM agents, and discusses techniques for enhancing the efficiency, memory manipulation, security, and privacy of Personal LLM Agents.

We posit that to achieve superhuman agents, future models require superhuman feedback in order to provide an adequate training signal. Current approaches commonly train reward models from human preferences, which may then be bottlenecked by human performance level, and secondly these separate frozen reward models cannot then learn to improve during LLM training. In this work, we study Self-Rewarding Language Models, where the language model itself is used via LLM-as-a-Judge prompting to provide its own rewards during training. We show that during Iterative DPO training that not only does instruction following ability improve, but also the ability to provide high-quality rewards to itself. Fine-tuning Llama 2 70B on three iterations of our approach yields a model that outperforms many existing systems on the AlpacaEval 2.0 leaderboard, including Claude 2, Gemini Pro, and GPT-4 0613. While only a preliminary study, this work opens the door to the possibility of models that can continually improve in both axes. 2024: Weizhe Yuan, Richard Yuanzhe Pang, Kyunghyun Cho, Sainbayar Sukhbaatar, Jing Xu, Jason Weston https://arxiv.org/pdf/2401.10020v1.pdf

Recently the state space models (SSMs) with efficient hardware-aware designs, i.e., Mamba, have shown great potential for long sequence modeling. Building efficient and generic vision backbones purely upon SSMs is an appealing direction. However, representing visual data is challenging for SSMs due to the position-sensitivity of visual data and the requirement of global context for visual understanding. In this paper, we show that the reliance of visual representation learning on self-attention is not necessary and propose a new generic vision backbone with bidirectional Mamba blocks (Vim), which marks the image sequences with position embeddings and compresses the visual representation with bidirectional state space models. On ImageNet classification, COCO object detection, and ADE20k semantic segmentation tasks, Vim achieves higher performance compared to well-established vision transformers like DeiT, while also demonstrating significantly improved computation&memory efficiency. For example, Vim is 2.8$\times$ faster than DeiT and saves 86.8% GPU memory when performing batch inference to extract features on images with a resolution of 1248$\times$1248. The results demonstrate that Vim is capable of overcoming the computation&memory constraints on performing Transformer-style understanding for high-resolution images and it has great potential to become the next-generation backbone for vision foundation models. Code is available at https://github.com/hustvl/Vim. 2024: Lianghui Zhu, Bencheng Liao, Qian Zhang, Xinlong Wang, Wenyu Liu, Xinggang Wang https://arxiv.org/pdf/2401.09417v1.pdf

A podcast discusses a new approach to code generation called AlphaCodium. It improves performance on code problems by using a test-based, multi-stage, code-oriented iterative flow. AlphaCodium significantly increases accuracy compared to GPT-4. The podcast also covers design concepts, tricks, and best practices for code generation, as well as comparing AlphaCodium to other methods in the literature.

As large language models (LLMs) are adopted as a fundamental component of language technologies, it is crucial to accurately characterize their performance. Because choices in prompt design can strongly influence model behavior, this design process is critical in effectively using any modern pre-trained generative language model. In this work, we focus on LLM sensitivity to a quintessential class of meaning-preserving design choices: prompt formatting. We find that several widely used open-source LLMs are extremely sensitive to subtle changes in prompt formatting in few-shot settings, with performance differences of up to 76 accuracy points when evaluated using LLaMA-2-13B. Sensitivity remains even when increasing model size, the number of few-shot examples, or performing instruction tuning. Our analysis suggests that work evaluating LLMs with prompting-based methods would benefit from reporting a range of performance across plausible prompt formats, instead of the currently-standard practice of reporting performance on a single format. We also show that format performance only weakly correlates between models, which puts into question the methodological validity of comparing models with an arbitrarily chosen, fixed prompt format. To facilitate systematic analysis we propose FormatSpread, an algorithm that rapidly evaluates a sampled set of plausible prompt formats for a given task, and reports the interval of expected performance without accessing model weights. Furthermore, we present a suite of analyses that characterize the nature of this sensitivity, including exploring the influence of particular atomic perturbations and the internal representation of particular formats. 2023: Melanie Sclar, Yejin Choi, Yulia Tsvetkov, Alane Suhr https://arxiv.org/pdf/2310.11324.pdf

This work elicits LLMs' inherent ability to handle long contexts without fine-tuning. The limited length of the training sequence during training may limit the application of Large Language Models (LLMs) on long input sequences for inference. In this work, we argue that existing LLMs themselves have inherent capabilities for handling long contexts. Based on this argument, we suggest extending LLMs' context window by themselves to fully utilize the inherent ability.We propose Self-Extend to stimulate LLMs' long context handling potential. The basic idea is to construct bi-level attention information: the group level and the neighbor level. The two levels are computed by the original model's self-attention, which means the proposed does not require any training. With only four lines of code modification, the proposed method can effortlessly extend existing LLMs' context window without any fine-tuning. We conduct comprehensive experiments and the results show that the proposed method can effectively extend existing LLMs' context window's length. 2024: Hongye Jin, Xiaotian Han, Jingfeng Yang, Zhimeng Jiang, Zirui Liu, Chia-yuan Chang, Huiyuan Chen, Xia Hu https://arxiv.org/pdf/2401.01325.pdf

Information extraction (IE) aims to extract structural knowledge (such as entities, relations, and events) from plain natural language texts. Recently, generative Large Language Models (LLMs) have demonstrated remarkable capabilities in text understanding and generation, allowing for generalization across various domains and tasks. As a result, numerous works have been proposed to harness abilities of LLMs and offer viable solutions for IE tasks based on a generative paradigm. To conduct a comprehensive systematic review and exploration of LLM efforts for IE tasks, in this study, we survey the most recent advancements in this field. We first present an extensive overview by categorizing these works in terms of various IE subtasks and learning paradigms, then we empirically analyze the most advanced methods and discover the emerging trend of IE tasks with LLMs. Based on thorough review conducted, we identify several insights in technique and promising research directions that deserve further exploration in future studies. We maintain a public repository and consistently update related resources at: \url{https://github.com/quqxui/Awesome-LLM4IE-Papers}. 2023: Derong Xu, Wei Chen, Wenjun Peng, Chao Zhang, Tong Xu, Xiangyu Zhao, Xian Wu, Yefeng Zheng, Enhong Chen https://arxiv.org/pdf/2312.17617v1.pdf

Explore the innovative DeepSeekMoE architecture and its strategies for achieving expert specialization in language models. Discover how finely segmented experts and shared knowledge can enhance efficiency while cutting computational costs. The conversation dives into performance benchmarks, revealing how DeepSeekMoE 2B rivals larger models with fewer parameters. Additionally, hear about the launch of the DeepSeek Mo16B model, designed for high-memory single GPU use, marking a significant step forward in language modeling research.

The utilization of long contexts poses a big challenge for large language models due to their limited context window length. Although the context window can be extended through fine-tuning, it will result in a considerable cost at both training and inference time, and exert an unfavorable impact to the LLM's original capabilities. In this work, we propose Activation Beacon, which condenses LLM's raw activations into more compact forms such that it can perceive a much longer context with a limited context window. Activation Beacon is introduced as a plug-and-play module for the LLM. It fully preserves the LLM's original capability on short contexts while extending the new capability on processing longer contexts. Besides, it works with short sliding windows to process the long context, which achieves a competitive memory and time efficiency in both training and inference. Activation Beacon is learned by the auto-regression task conditioned on a mixture of beacons with diversified condensing ratios. Thanks to such a treatment, it can be efficiently trained purely with short-sequence data in just 10K steps, which consumes less than 9 hours on a single 8xA800 GPU machine. The experimental studies show that Activation Beacon is able to extend Llama-2-7B's context length by $\times100$ times (from 4K to 400K), meanwhile achieving a superior result on both long-context generation and understanding tasks. Our model and code will be available at the BGE repository. 2024: Peitian Zhang, Zheng Liu, Shitao Xiao, Ninglu Shao, Qiwei Ye, Zhicheng Dou https://arxiv.org/pdf/2401.03462v1.pdf

Fine-tuning large pre-trained models is an effective transfer mechanism in NLP. However, in the presence of many downstream tasks, fine-tuning is parameter inefficient: an entire new model is required for every task. As an alternative, we propose transfer with adapter modules. Adapter modules yield a compact and extensible model; they add only a few trainable parameters per task, and new tasks can be added without revisiting previous ones. The parameters of the original network remain fixed, yielding a high degree of parameter sharing. To demonstrate adapter's effectiveness, we transfer the recently proposed BERT Transformer model to 26 diverse text classification tasks, including the GLUE benchmark. Adapters attain near state-of-the-art performance, whilst adding only a few parameters per task. On GLUE, we attain within 0.4% of the performance of full fine-tuning, adding only 3.6% parameters per task. By contrast, fine-tuning trains 100% of the parameters per task. 2019: N. Houlsby, A. Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin de Laroussilhe, Andrea Gesmundo, Mona Attariyan, S. Gelly Natural language processing, Benchmark (computing), Transformer, Document classification, Downstream (software development), While https://arxiv.org/pdf/1902.00751v1.pdf