Scaling laws with respect to the amount of training data and the number of parameters allow us to predict the cost-benefit trade-offs of pretraining language models (LMs) in different configurations. In this paper, we consider another dimension of scaling: the amount of data available at inference time. Specifically, we find that increasing the size of the datastore used by a retrieval-based LM monotonically improves language modeling and several downstream tasks without obvious saturation, such that a smaller model augmented with a large datastore outperforms a larger LM-only model on knowledge-intensive tasks. By plotting compute-optimal scaling curves with varied datastore, model, and pretraining data sizes, we show that using larger datastores can significantly improve model performance for the same training compute budget. We carry out our study by constructing a 1.4 trillion-token datastore named MassiveDS, which is the largest and the most diverse open-sourced datastore for retrieval-based LMs to date, and designing an efficient pipeline for studying datastore scaling in a computationally accessible manner. Finally, we analyze the effect of improving the retriever, datastore quality filtering, and other design choices on our observed scaling trends. Overall, our results show that datastore size should be considered as an integral part of LM efficiency and performance trade-offs. To facilitate future research, we open-source our datastore and code at https://github.com/RulinShao/retrieval-scaling.

Optimization problems seek to find the best solution to an objective under a set of constraints, and have been widely investigated in real-world applications. Modeling and solving optimization problems in a specific domain typically require a combination of domain knowledge, mathematical skills, and programming ability, making it difficult for general users and even domain professionals. In this paper, we propose a novel framework called OptLLM that augments LLMs with external solvers. Specifically, OptLLM accepts user queries in natural language, convert them into mathematical formulations and programming codes, and calls the solvers to calculate the results for decision-making. In addition, OptLLM supports multi-round dialogues to gradually refine the modeling and solving of optimization problems. To illustrate the effectiveness of OptLLM, we provide tutorials on three typical optimization applications and conduct experiments on both prompt-based GPT models and a fine-tuned Qwen model using a large-scale selfdeveloped optimization dataset. Experimental results show that OptLLM works with various LLMs, and the fine-tuned model achieves an accuracy boost compared to the promptbased models. Some features of OptLLM framework have been available for trial since June 2023 (https://opt.alibabacloud.com/chat or https://opt.aliyun.com/chat).

We introduce WildGuard -- an open, light-weight moderation tool for LLM safety that achieves three goals: (1) identifying malicious intent in user prompts, (2) detecting safety risks of model responses, and (3) determining model refusal rate. Together, WildGuard serves the increasing needs for automatic safety moderation and evaluation of LLM interactions, providing a one-stop tool with enhanced accuracy and broad coverage across 13 risk categories. While existing open moderation tools such as Llama-Guard2 score reasonably well in classifying straightforward model interactions, they lag far behind a prompted GPT-4, especially in identifying adversarial jailbreaks and in evaluating models' refusals, a key measure for evaluating safety behaviors in model responses. To address these challenges, we construct WildGuardMix, a large-scale and carefully balanced multi-task safety moderation dataset with 92K labeled examples that cover vanilla (direct) prompts and adversarial jailbreaks, paired with various refusal and compliance responses. WildGuardMix is a combination of WildGuardTrain, the training data of WildGuard, and WildGuardTest, a high-quality human-annotated moderation test set with 5K labeled items covering broad risk scenarios. Through extensive evaluations on WildGuardTest and ten existing public benchmarks, we show that WildGuard establishes state-of-the-art performance in open-source safety moderation across all the three tasks compared to ten strong existing open-source moderation models (e.g., up to 26.4% improvement on refusal detection). Importantly, WildGuard matches and sometimes exceeds GPT-4 performance (e.g., up to 3.9% improvement on prompt harmfulness identification). WildGuard serves as a highly effective safety moderator in an LLM interface, reducing the success rate of jailbreak attacks from 79.8% to 2.4%.

As Large language models have shown a remarkable a significant milestone in this area, Elhage et al. ability to learn and perform complex tasks through (2021) demonstrated the existence of induction in-context learning (ICL) (Brown et al., 2020; Touvron heads in Transformer LMs. These heads scan the et al., 2023b). In ICL, the model receives context for previous instances of the current token a demonstration context and a query question as using a prefix matching mechanism, which identifies a prompt for prediction. Unlike supervised learning, if and where a token has appeared before. ICL utilises the pretrained model's capabilities If a matching token is found, the head employs to recognise and replicate patterns within the a copying mechanism to increase the probability demonstration context, thereby enabling accurate of the subsequent token, facilitating exact or approximate predictions for the query without the use of gradient repetition of sequences and embodying updates.

Large language models (LLMs) are increasingly used in robotics, especially for high-level action planning. Meanwhile, many robotics applications involve human supervisors or collaborators. Hence, it is crucial for LLMs to generate socially acceptable actions that align with people's preferences and values. In this work, we test whether LLMs capture people's intuitions about behavior judgments and communication preferences in human-robot interaction (HRI) scenarios. For evaluation, we reproduce three HRI user studies, comparing the output of LLMs with that of real participants. We find that GPT-4 strongly outperforms other models, generating answers that correlate strongly with users' answers in two studies $\unicode{x2014}$ the first study dealing with selecting the most appropriate communicative act for a robot in various situations ($r_s$ = 0.82), and the second with judging the desirability, intentionality, and surprisingness of behavior ($r_s$ = 0.83). However, for the last study, testing whether people judge the behavior of robots and humans differently, no model achieves strong correlations. Moreover, we show that vision models fail to capture the essence of video stimuli and that LLMs tend to rate different communicative acts and behavior desirability higher than people.

Humans describe complex scenes with compositionality, using simple text descriptions enriched with links and relationships. While vision-language research has aimed to develop models with compositional understanding capabilities, this is not reflected yet in existing datasets which, for the most part, still use plain text to describe images. In this work, we propose a new annotation strategy, graph-based captioning (GBC) that describes an image using a labelled graph structure, with nodes of various types. The nodes in GBC are created using, in a first stage, object detection and dense captioning tools nested recursively to uncover and describe entity nodes, further linked together in a second stage by highlighting, using new types of nodes, compositions and relations among entities. Since all GBC nodes hold plain text descriptions, GBC retains the flexibility found in natural language, but can also encode hierarchical information in its edges. We demonstrate that GBC can be produced automatically, using off-the-shelf multimodal LLMs and open-vocabulary detection models, by building a new dataset, GBC10M, gathering GBC annotations for about 10M images of the CC12M dataset. We use GBC10M to showcase the wealth of node captions uncovered by GBC, as measured with CLIP training. We show that using GBC nodes' annotations -- notably those stored in composition and relation nodes -- results in significant performance boost on downstream models when compared to other dataset formats. To further explore the opportunities provided by GBC, we also propose a new attention mechanism that can leverage the entire GBC graph, with encouraging experimental results that show the extra benefits of incorporating the graph structure. Our datasets are released at \url{https://huggingface.co/graph-based-captions}.

Speech Integrated Large Language Models (SILLMs) combine large language models with speech perception to perform diverse tasks, such as emotion recognition to speaker verification, demonstrating universal audio understanding capability. However, these models may amplify biases present in training data, potentially leading to biased access to information for marginalized groups. This work introduces a curated spoken bias evaluation toolkit and corresponding dataset. We evaluate gender bias in SILLMs across four semantic-related tasks: speech-to-text translation (STT), spoken coreference resolution (SCR), spoken sentence continuation (SSC), and spoken question answering (SQA). Our analysis reveals that bias levels are language-dependent and vary with different evaluation methods. Our findings emphasize the necessity of employing multiple approaches to comprehensively assess biases in SILLMs, providing insights for developing fairer SILLM systems.

Large language models (LLMs) have shown exceptional performance and vast potential across diverse tasks. However, the deployment of LLMs with high performance in low-resource environments has garnered significant attention in the industry. When GPU hardware resources are limited, we can explore alternative options on CPUs. To mitigate the financial burden and alleviate constraints imposed by hardware resources, optimizing inference performance is necessary. In this paper, we introduce an easily deployable inference performance optimization solution aimed at accelerating LLMs on CPUs. In this solution, we implement an effective way to reduce the KV cache size while ensuring precision. We propose a distributed inference optimization approach and implement it based on oneAPI Collective Communications Library. Furthermore, we propose optimization approaches for LLMs on CPU, and conduct tailored optimizations for the most commonly used models. The code is open-sourced at https://github.com/intel/xFasterTransformer.

Generative pre-trained transformers (GPT's) are a type of large language machine learning model that are unusually adept at producing novel, and coherent, natural language. In this study the ability of GPT models to generate novel and correct versions, and notably very insecure versions, of implementations of the cryptographic hash function SHA-1 is examined. The GPT models Llama-2-70b-chat-h, Mistral-7B-Instruct-v0.1, and zephyr-7b-alpha are used. The GPT models are prompted to re-write each function using a modified version of the localGPT framework and langchain to provide word embedding context of the full source code and header files to the model, resulting in over 150,000 function re-write GPT output text blocks, approximately 50,000 of which were able to be parsed as C code and subsequently compiled. The generated code is analyzed for being compilable, correctness of the algorithm, memory leaks, compiler optimization stability, and character distance to the reference implementation. Remarkably, several generated function variants have a high implementation security risk of being correct for some test vectors, but incorrect for other test vectors. Additionally, many function implementations were not correct to the reference algorithm of SHA-1, but produced hashes that have some of the basic characteristics of hash functions. Many of the function re-writes contained serious flaws such as memory leaks, integer overflows, out of bounds accesses, use of uninitialised values, and compiler optimization instability. Compiler optimization settings and SHA-256 hash checksums of the compiled binaries are used to cluster implementations that are equivalent but may not have identical syntax - using this clustering over 100,000 novel and correct versions of the SHA-1 codebase were generated where each component C function of the reference implementation is different from the original code.

Large Language Models (LLMs) have been applied to many research problems across various domains. One of the applications of LLMs is providing question-answering systems that cater to users from different fields. The effectiveness of LLM-based question-answering systems has already been established at an acceptable level for users posing questions in popular and public domains such as trivia and literature. However, it has not often been established in niche domains that traditionally require specialized expertise. To this end, we construct the NEPAQuAD1.0 benchmark to evaluate the performance of three frontier LLMs -- Claude Sonnet, Gemini, and GPT-4 -- when answering questions originating from Environmental Impact Statements prepared by U.S. federal government agencies in accordance with the National Environmental Environmental Act (NEPA). We specifically measure the ability of LLMs to understand the nuances of legal, technical, and compliance-related information present in NEPA documents in different contextual scenarios. For example, we test the LLMs' internal prior NEPA knowledge by providing questions without any context, as well as assess how LLMs synthesize the contextual information present in long NEPA documents to facilitate the question/answering task. We compare the performance of the long context LLMs and RAG powered models in handling different types of questions (e.g., problem-solving, divergent). Our results suggest that RAG powered models significantly outperform the long context models in the answer accuracy regardless of the choice of the frontier LLM. Our further analysis reveals that many models perform better answering closed questions than divergent and problem-solving questions.