The traditional notion of "Junk DNA" has long been linked to non-coding segments within the human genome, constituting roughly 98% of its composition. However, recent research has unveiled the critical roles some of these seemingly non-functional DNA sequences play in cellular processes. Intriguingly, the weights within deep neural networks exhibit a remarkable similarity to the redundancy observed in human genes. It was believed that weights in gigantic models contained excessive redundancy, and could be removed without compromising performance. This paper challenges this conventional wisdom by presenting a compelling counter-argument. We employ sparsity as a tool to isolate and quantify the nuanced significance of low-magnitude weights in pre-trained large language models (LLMs). Our study demonstrates a strong correlation between these weight magnitudes and the knowledge they encapsulate, from a downstream task-centric angle. we raise the "Junk DNA Hypothesis" backed by our in-depth investigation: while small-magnitude weights may appear "useless" for simple tasks and suitable for pruning, they actually encode crucial knowledge necessary for solving more difficult downstream tasks. Removing these seemingly insignificant weights can lead to irreversible knowledge forgetting and performance damage in difficult tasks. These findings offer fresh insights into how LLMs encode knowledge in a task-sensitive manner, pave future research direction in model pruning, and open avenues for task-aware conditional computation during inference.

Classifiers built upon vision-language models such as CLIP have shown remarkable zero-shot performance across a broad range of image classification tasks. Prior work has studied different ways of automatically creating descriptor sets for every class based on prompt templates, ranging from manually engineered templates over templates obtained from a large language model to templates built from random words and characters. Up until now, deriving zero-shot classifiers from the respective encoded class descriptors has remained nearly unchanged, i.e., classify to the class that maximizes cosine similarity between its averaged encoded class descriptors and the image encoding. However, weighing all class descriptors equally can be suboptimal when certain descriptors match visual clues on a given image better than others. In this work, we propose AutoCLIP, a method for auto-tuning zero-shot classifiers. AutoCLIP tunes per-image weights to each prompt template at inference time, based on statistics of class descriptor-image similarities. AutoCLIP is fully unsupervised, has very low computational overhead, and can be easily implemented in few lines of code. We show that AutoCLIP outperforms baselines across a broad range of vision-language models, datasets, and prompt templates consistently and by up to 3 percent point accuracy.

We demonstrate that Contrastive Decoding - a simple, computationally light, and training-free text generation method proposed by Li et al 2022 - achieves large out-of-the-box improvements over greedy decoding on a variety of reasoning tasks. Originally shown to improve the perceived quality of long-form text generation, Contrastive Decoding searches for strings that maximize a weighted difference in likelihood between strong and weak models. We show that Contrastive Decoding leads LLaMA-65B to outperform LLaMA 2, GPT-3.5 and PaLM 2-L on the HellaSwag commonsense reasoning benchmark, and to outperform LLaMA 2, GPT-3.5 and PaLM-540B on the GSM8K math word reasoning benchmark, in addition to improvements on a collection of other tasks. Analysis suggests that Contrastive Decoding improves over existing methods by preventing some abstract reasoning errors, as well as by avoiding simpler modes such as copying sections of the input during chain-of-thought. Overall, Contrastive Decoding outperforms nucleus sampling for long-form generation and greedy decoding for reasoning tasks, making it a powerful general purpose method for generating text from language models. Figure 1: Contrastive decoding improves reasoning Figure 2: Contrastive scoring significantly improves across model scales and reasoning tasks. Text is generated from large language models (LLMs) in different ways for different tasks. For openended text generation tasks, truncated sampling is normally used, as the most likely strings under a model tend to be short and uninteresting (Holtzman et al., 2020). For reasoning problems, greedy decoding is normally preferred, to avoid risking sampling errors.

In today's critical domains, the predominance of black-box machine learning models amplifies the demand for Explainable AI (XAI). The widely used SHAP values, while quantifying feature importance, are often too intricate and lack human-friendly explanations. Furthermore, counterfactual (CF) explanations present `what ifs' but leave users grappling with the 'why'. To bridge this gap, we introduce XAIstories. Leveraging Large Language Models, XAIstories provide narratives that shed light on AI predictions: SHAPstories do so based on SHAP explanations to explain a prediction score, while CFstories do so for CF explanations to explain a decision. Our results are striking: over 90% of the surveyed general audience finds the narrative generated by SHAPstories convincing. Data scientists primarily see the value of SHAPstories in communicating explanations to a general audience, with 92% of data scientists indicating that it will contribute to the ease and confidence of nonspecialists in understanding AI predictions. Additionally, 83% of data scientists indicate they are likely to use SHAPstories for this purpose. In image classification, CFstories are considered more or equally convincing as users own crafted stories by over 75% of lay user participants. CFstories also bring a tenfold speed gain in creating a narrative, and improves accuracy by over 20% compared to manually created narratives. The results thereby suggest that XAIstories may provide the missing link in truly explaining and understanding AI predictions.

Recent decisions by leading AI labs to either open-source their models or to restrict access to their models has sparked debate about whether, and how, increasingly capable AI models should be shared. Open-sourcing in AI typically refers to making model architecture and weights freely and publicly accessible for anyone to modify, study, build on, and use. This offers advantages such as enabling external oversight, accelerating progress, and decentralizing control over AI development and use. However, it also presents a growing potential for misuse and unintended consequences. This paper offers an examination of the risks and benefits of open-sourcing highly capable foundation models. While open-sourcing has historically provided substantial net benefits for most software and AI development processes, we argue that for some highly capable foundation models likely to be developed in the near future, open-sourcing may pose sufficiently extreme risks to outweigh the benefits. In such a case, highly capable foundation models should not be open-sourced, at least not initially. Alternative strategies, including non-open-source model sharing options, are explored. The paper concludes with recommendations for developers, standard-setting bodies, and governments for establishing safe and responsible model sharing practices and preserving open-source benefits where safe.

The field of Artificial Intelligence has witnessed remarkable progress in recent years, especially with the emergence of powerful large language models (LLMs) based on the transformer architecture. Cloud-based LLMs, such as OpenAI's ChatGPT, offer impressive capabilities but come with concerns regarding latency and privacy due to network dependencies. This article presents an innovative approach to LLM inference, envisioning a future where LLMs with billions of parameters can be executed directly on mobile devices without network connectivity. The article showcases a fine-tuned GPT LLM with 3 billion parameters that can operate smoothly on devices with as low as 4GB of memory. Through the integration of native code and model quantization techniques, the application not only serves as a general-purpose assistant but also facilitates seamless mobile interactions with text-to-actions features. The article provides insights into the training pipeline, implementation details, test results, and future directions of on-device LLM inference. This breakthrough technology opens up possibilities for empowering users with sophisticated AI capabilities while preserving their privacy and eliminating latency concerns.

Large language models (LLMs) demonstrate impressive performance on a wide variety of tasks, but they often struggle with tasks that require multi-step reasoning or goal-directed planning. To address this, we take inspiration from the human brain, in which planning is accomplished via the recurrent interaction of specialized modules in the prefrontal cortex (PFC). These modules perform functions such as conflict monitoring, state prediction, state evaluation, task decomposition, and task coordination. We find that LLMs are sometimes capable of carrying out these functions in isolation, but struggle to autonomously coordinate them in the service of a goal. Therefore, we propose a black box architecture with multiple LLM-based (GPT-4) modules. The architecture improves planning through the interaction of specialized PFC-inspired modules that break down a larger problem into multiple brief automated calls to the LLM. We evaluate the combined architecture on two challenging planning tasks -- graph traversal and Tower of Hanoi -- finding that it yields significant improvements over standard LLM methods (e.g., zero-shot prompting or in-context learning). These results demonstrate the benefit of utilizing knowledge from cognitive neuroscience to improve planning in LLMs.

Exploring rich environments and evaluating one's actions without prior knowledge is immensely challenging. In this paper, we propose Motif, a general method to interface such prior knowledge from a Large Language Model (LLM) with an agent. Motif is based on the idea of grounding LLMs for decision-making without requiring them to interact with the environment: it elicits preferences from an LLM over pairs of captions to construct an intrinsic reward, which is then used to train agents with reinforcement learning. We evaluate Motif's performance and behavior on the challenging, open-ended and procedurally-generated NetHack game. Surprisingly, by only learning to maximize its intrinsic reward, Motif achieves a higher game score than an algorithm directly trained to maximize the score itself. When combining Motif's intrinsic reward with the environment reward, our method significantly outperforms existing approaches and makes progress on tasks where no advancements have ever been made without demonstrations. Finally, we show that Motif mostly generates intuitive human-aligned behaviors which can be steered easily through prompt modifications, while scaling well with the LLM size and the amount of information given in the prompt.

Cooking recipes are especially challenging to translate to robot plans as they feature rich linguistic complexity, temporally-extended interconnected tasks, and an almost infinite space of possible actions. Our key insight is that combining a source of background cooking domain knowledge with a formalism capable of handling the temporal richness of cooking recipes could enable the extraction of unambiguous, robot-executable plans. In this work, we use Linear Temporal Logic (LTL) as a formal language expressible enough to model the temporal nature of cooking recipes. Leveraging pre-trained Large Language Models (LLMs), we present a system that translates instruction steps from an arbitrary cooking recipe found on the internet to a series of LTL formulae, grounding high-level cooking actions to a set of primitive actions that are executable by a manipulator in a kitchen environment. Our approach makes use of a caching scheme, dynamically building a queryable action library at runtime, significantly decreasing LLM API calls (-51%), latency (-59%) and cost (-42%) compared to a baseline that queries the LLM for every newly encountered action at runtime. We demonstrate the transferability of our system in a realistic simulation platform through showcasing a set of simple cooking tasks.

Recent works have demonstrated the effectiveness of self-alignment in which a large language model is, by itself, aligned to follow general instructions through the automatic generation of instructional data using a handful of human-written seeds. Instead of general alignment, in this work, we focus on self-alignment for expert domain specialization (e.g., biomedicine), discovering it to be very effective for improving zero-shot and few-shot performance in target domains of interest. As a preliminary, we first present the benchmark results of existing aligned models within a specialized domain, which reveals the marginal effect that "generic" instruction-following training has on downstream expert domains' performance. To remedy this, we explore self-specialization that leverages domain-specific unlabelled data and a few labeled seeds for the self-alignment process. When augmented with retrieval to reduce hallucination and enhance concurrency of the alignment, self-specialization offers an effective (and efficient) way of "carving out" an expert model out of a "generalist", pre-trained LLM where different domains of expertise are originally combined in a form of "superposition". Our experimental results on a biomedical domain show that our self-specialized model (30B) outperforms its base model, MPT-30B by a large margin and even surpasses larger popular models based on LLaMA-65B, highlighting its potential and practicality for specialization, especially considering its efficiency in terms of data and parameters.