Can LLMs Reason in the Wild with Programs?

Large Language Models (LLMs) have shown superior capability to solve reasoning problems with programs. While being a promising direction, most of such frameworks are trained and evaluated in settings with a prior knowledge of task requirements. However, as LLMs become more capable, it is necessary to assess their reasoning abilities in more realistic scenarios where many real-world problems are open-ended with ambiguous scope, and often require multiple formalisms to solve. To investigate this, we introduce the task of reasoning in the wild, where an LLM is tasked to solve a reasoning problem of unknown type by identifying the subproblems and their corresponding formalisms, and writing a program to solve each subproblem, guided by a tactic. We create a large tactic-guided trajectory dataset containing detailed solutions to a diverse set of reasoning problems, ranging from well-defined single-form reasoning (e.g., math, logic), to ambiguous and hybrid ones (e.g., commonsense, combined math and logic). This allows us to test various aspects of LLMs reasoning at the fine-grained level such as the selection and execution of tactics, and the tendency to take undesired shortcuts. In experiments, we highlight that existing LLMs fail significantly on problems with ambiguous and mixed scope, revealing critical limitations and overfitting issues (e.g. accuracy on GSM8K drops by at least 50\%). We further show the potential of finetuning a local LLM on the tactic-guided trajectories in achieving better performance. Project repo is available at github.com/gblackout/Reason-in-the-Wild

Towards Hierarchical Multi-Agent Workflows for Zero-Shot Prompt Optimization

Large language models (LLMs) have shown great progress in responding to user questions, allowing for a multitude of diverse applications. Yet, the quality of LLM outputs heavily depends on the prompt design, where a good prompt might enable the LLM to answer a very challenging question correctly. Therefore, recent works have developed many strategies for improving the prompt, including both manual crafting and in-domain optimization. However, their efficacy in unrestricted scenarios remains questionable, as the former depends on human design for specific questions and the latter usually generalizes poorly to unseen scenarios. To address these problems, we give LLMs the freedom to design the best prompts according to themselves. Specifically, we include a hierarchy of LLMs, first constructing a prompt with precise instructions and accurate wording in a hierarchical manner, and then using this prompt to generate the final answer to the user query. We term this pipeline Hierarchical Multi-Agent Workflow, or HMAW. In contrast with prior works, HMAW imposes no human restriction and requires no training, and is completely task-agnostic while capable of adjusting to the nuances of the underlying task. Through both quantitative and qualitative experiments across multiple benchmarks, we verify that despite its simplicity, the proposed approach can create detailed and suitable prompts, further boosting the performance of current LLMs.

LMO-DP: Optimizing the Randomization Mechanism for Differentially Private Fine-Tuning (Large) Language Models

Differentially Private Stochastic Gradient Descent (DP-SGD) and its variants have been proposed to ensure rigorous privacy for fine-tuning large-scale pre-trained language models. However, they rely heavily on the Gaussian mechanism, which may overly perturb the gradients and degrade the accuracy, especially in stronger privacy regimes (e.g., the privacy budget $\epsilon < 3$). To address such limitations, we propose a novel Language Model-based Optimal Differential Privacy (LMO-DP) mechanism, which takes the first step to enable the tight composition of accurately fine-tuning (large) language models with a sub-optimal DP mechanism, even in strong privacy regimes (e.g., $0.1\leq \epsilon<3$). Furthermore, we propose a novel offline optimal noise search method to efficiently derive the sub-optimal DP that significantly reduces the noise magnitude. For instance, fine-tuning RoBERTa-large (with 300M parameters) on the SST-2 dataset can achieve an accuracy of 92.20% (given $\epsilon=0.3$, $\delta=10^{-10}$) by drastically outperforming the Gaussian mechanism (e.g., $\sim 50\%$ for small $\epsilon$ and $\delta$). We also draw similar findings on the text generation tasks on GPT-2. Finally, to our best knowledge, LMO-DP is also the first solution to accurately fine-tune Llama-2 with strong differential privacy guarantees. The code will be released soon and available upon request.

SketchQL Demonstration: Zero-shot Video Moment Querying with Sketches

In this paper, we will present SketchQL, a video database management system (VDBMS) for retrieving video moments with a sketch-based query interface. This novel interface allows users to specify object trajectory events with simple mouse drag-and-drop operations. Users can use trajectories of single objects as building blocks to compose complex events. Using a pre-trained model that encodes trajectory similarity, SketchQL achieves zero-shot video moments retrieval by performing similarity searches over the video to identify clips that are the most similar to the visual query. In this demonstration, we introduce the graphic user interface of SketchQL and detail its functionalities and interaction mechanisms. We also demonstrate the end-to-end usage of SketchQL from query composition to video moments retrieval using real-world scenarios.

Generalization Error Bounds for Learning under Censored Feedback

Generalization error bounds from learning theory provide statistical guarantees on how well an algorithm will perform on previously unseen data. In this paper, we characterize the impacts of data non-IIDness due to censored feedback (a.k.a. selective labeling bias) on such bounds. We first derive an extension of the well-known Dvoretzky-Kiefer-Wolfowitz (DKW) inequality, which characterizes the gap between empirical and theoretical CDFs given IID data, to problems with non-IID data due to censored feedback. We then use this CDF error bound to provide a bound on the generalization error guarantees of a classifier trained on such non-IID data. We show that existing generalization error bounds (which do not account for censored feedback) fail to correctly capture the model's generalization guarantees, verifying the need for our bounds. We further analyze the effectiveness of (pure and bounded) exploration techniques, proposed by recent literature as a way to alleviate censored feedback, on improving our error bounds. Together, our findings illustrate how a decision maker should account for the trade-off between strengthening the generalization guarantees of an algorithm and the costs incurred in data collection when future data availability is limited by censored feedback.

Not All Federated Learning Algorithms Are Created Equal: A Performance Evaluation Study

Federated Learning (FL) emerged as a practical approach to training a model from decentralized data. The proliferation of FL led to the development of numerous FL algorithms and mechanisms. Many prior efforts have given their primary focus on accuracy of those approaches, but there exists little understanding of other aspects such as computational overheads, performance and training stability, etc. To bridge this gap, we conduct extensive performance evaluation on several canonical FL algorithms (FedAvg, FedProx, FedYogi, FedAdam, SCAFFOLD, and FedDyn) by leveraging an open-source federated learning framework called Flame. Our comprehensive measurement study reveals that no single algorithm works best across different performance metrics. A few key observations are: (1) While some state-of-the-art algorithms achieve higher accuracy than others, they incur either higher computation overheads (FedDyn) or communication overheads (SCAFFOLD). (2) Recent algorithms present smaller standard deviation in accuracy across clients than FedAvg, indicating that the advanced algorithms' performances are stable. (3) However, algorithms such as FedDyn and SCAFFOLD are more prone to catastrophic failures without the support of additional techniques such as gradient clipping. We hope that our empirical study can help the community to build best practices in evaluating FL algorithms.

Data-Efficient Contrastive Language-Image Pretraining: Prioritizing Data Quality over Quantity

Contrastive Language-Image Pre-training (CLIP) on large-scale image-caption datasets learns representations that can achieve remarkable zero-shot generalization. However, such models require a massive amount of pre-training data. Improving the quality of the pre-training data has been shown to be much more effective in improving CLIP's performance than increasing its volume. Nevertheless, finding small subsets of training data that provably generalize the best has remained an open question. In this work, we propose the first theoretically rigorous data selection method for CLIP. We show that subsets that closely preserve the cross-covariance of the images and captions of the full data provably achieve a superior generalization performance. Our extensive experiments on ConceptualCaptions3M and ConceptualCaptions12M demonstrate that subsets found by \method\ achieve over 2.7x and 1.4x the accuracy of the next best baseline on ImageNet and its shifted versions. Moreover, we show that our subsets obtain 1.5x the average accuracy across 11 downstream datasets, of the next best baseline. The code is available at: https://github.com/BigML-CS-UCLA/clipcov-data-efficient-clip.

Prompt Mining for Language-based Human Mobility Forecasting

With the advancement of large language models, language-based forecasting has recently emerged as an innovative approach for predicting human mobility patterns. The core idea is to use prompts to transform the raw mobility data given as numerical values into natural language sentences so that the language models can be leveraged to generate the description for future observations. However, previous studies have only employed fixed and manually designed templates to transform numerical values into sentences. Since the forecasting performance of language models heavily relies on prompts, using fixed templates for prompting may limit the forecasting capability of language models. In this paper, we propose a novel framework for prompt mining in language-based mobility forecasting, aiming to explore diverse prompt design strategies. Specifically, the framework includes a prompt generation stage based on the information entropy of prompts and a prompt refinement stage to integrate mechanisms such as the chain of thought. Experimental results on real-world large-scale data demonstrate the superiority of generated prompts from our prompt mining pipeline. Additionally, the comparison of different prompt variants shows that the proposed prompt refinement process is effective. Our study presents a promising direction for further advancing language-based mobility forecasting.

When is Tree Search Useful for LLM Planning? It Depends on the Discriminator

In this paper, we examine how large language models (LLMs) solve multi-step problems under a language agent framework with three components: a generator, a discriminator, and a planning method. We investigate the practical utility of two advanced planning methods, iterative correction and tree search. We present a comprehensive analysis of how discrimination accuracy affects the overall performance of agents when using these two methods or a simpler method, re-ranking. Experiments on two tasks, text-to-SQL parsing and mathematical reasoning, show that: (1) advanced planning methods demand discriminators with at least 90% accuracy to achieve significant improvements over re-ranking; (2) current LLMs' discrimination abilities have not met the needs of advanced planning methods to achieve such improvements; (3) with LLM-based discriminators, advanced planning methods may not adequately balance accuracy and efficiency. For example, compared to the other two methods, tree search is at least 10--20 times slower but leads to negligible performance gains, which hinders its real-world applications. Code and data will be released at https://github.com/OSU-NLP-Group/llm-planning-eval.

Large Language Models Can Learn Temporal Reasoning

Large language models (LLMs) learn temporal concepts from the co-occurrence of related tokens in a sequence. Compared with conventional text generation, temporal reasoning, which reaches a conclusion based on mathematical, logical and commonsense knowledge, is more challenging. In this paper, we propose TempGraph-LLM, a new paradigm towards text-based temporal reasoning. To be specific, we first teach LLMs to translate the context into a temporal graph. A synthetic dataset, which is fully controllable and requires minimal supervision, is constructed for pre-training on this task. We prove in experiments that LLMs benefit from the pre-training on other tasks. On top of that, we guide LLMs to perform symbolic reasoning with the strategies of Chain of Thoughts (CoTs) bootstrapping and special data augmentation. We observe that CoTs with symbolic reasoning bring more consistent and reliable results than those using free text.