Search

Recently, deep generative models have greatly advanced the progress of face video coding towards promising rate-distortion performance and diverse application functionalities. Beyond traditional hybrid video coding paradigms, Generative Face Video Compression (GFVC) relying on the strong capabilities of deep generative models and the philosophy of early Model-Based Coding (MBC) can facilitate the compact representation and realistic reconstruction of visual face signal, thus achieving ultra-low bitrate face video communication. However, these GFVC algorithms are sometimes faced with unstable reconstruction quality and limited bitrate ranges. To address these problems, this paper proposes a novel Progressive Face Video Compression framework, namely PFVC, that utilizes adaptive visual tokens to realize exceptional trade-offs between reconstruction robustness and bandwidth intelligence. In particular, the encoder of the proposed PFVC projects the high-dimensional face signal into adaptive visual tokens in a progressive manner, whilst the decoder can further reconstruct these adaptive visual tokens for motion estimation and signal synthesis with different granularity levels. Experimental results demonstrate that the proposed PFVC framework can achieve better coding flexibility and superior rate-distortion performance in comparison with the latest Versatile Video Coding (VVC) codec and the state-of-the-art GFVC algorithms. The project page can be found at https://github.com/Berlin0610/PFVC.

Large Language Model (LLM) services exhibit impressive capability on unlearned tasks leveraging only a few examples by in-context learning (ICL). However, the success of ICL varies depending on the task and context, leading to heterogeneous service quality. Directly estimating the performance of LLM services at each invocation can be laborious, especially requiring abundant labeled data or internal information within the LLM. This paper introduces a novel method to estimate the performance of LLM services across different tasks and contexts, which can be "plug-and-play" utilizing only a few unlabeled samples like ICL. Our findings suggest that the negative log-likelihood and perplexity derived from LLM service invocation can function as effective and significant features. Based on these features, we utilize four distinct meta-models to estimate the performance of LLM services. Our proposed method is compared against unlabeled estimation baselines across multiple LLM services and tasks. And it is experimentally applied to two scenarios, demonstrating its effectiveness in the selection and further optimization of LLM services.

With the rapid advancement of large language models (LLMs), recent years have witnessed many promising studies on leveraging LLM-based agents to simulate human social behavior. While prior work has demonstrated significant potential across various domains, much of it has focused on specific scenarios involving a limited number of agents and has lacked the ability to adapt when errors occur during simulation. To overcome these limitations, we propose a novel LLM-agent-based simulation platform called \textit{GenSim}, which: (1) \textbf{Abstracts a set of general functions} to simplify the simulation of customized social scenarios; (2) \textbf{Supports one hundred thousand agents} to better simulate large-scale populations in real-world contexts; (3) \textbf{Incorporates error-correction mechanisms} to ensure more reliable and long-term simulations. To evaluate our platform, we assess both the efficiency of large-scale agent simulations and the effectiveness of the error-correction mechanisms. To our knowledge, GenSim represents an initial step toward a general, large-scale, and correctable social simulation platform based on LLM agents, promising to further advance the field of social science.

Through the collaboration of multiple agents possessing diverse expertise and tools, multi-agent systems achieve impressive progress in solving real-world problems. Given the user queries, the meta-agents, serving as the brain within these systems, are required to decompose the queries into multiple sub-tasks that can be allocated to suitable agents capable of solving them, so-called agent-oriented planning. In this study, we identify three critical design principles of agent-oriented planning, including solvability, completeness, and non-redundancy, to ensure that each sub-task is effectively resolved, leading to satisfactory responses to the original queries. These principles further inspire us to propose a novel framework for agent-oriented planning in multi-agent systems, leveraging a fast task decomposition and allocation process followed by an effective and efficient evaluation via a reward model. During the planning process, the meta-agent is also responsible for evaluating the performance of the expert agents, making timely adjustments to the sub-tasks and scheduling as necessary. Besides, we integrate a feedback loop into the proposed framework to further enhance the effectiveness and robustness of such a problem-solving process. Extensive experiments demonstrate the advancement of the proposed framework in solving real-world problems compared to both single-agent systems and existing planning strategies for multi-agent systems.

Near-sun sky twilight observations allow for the detection of asteroid interior to the orbit of Venus (Aylos), the Earth (Atiras), and comets. We present the results of observations with the Palomar 48-inch telescope (P48)/Zwicky Transient Facility (ZTF) camera in 30 s r-band exposures taken during evening astronomical twilight from 2019 Sep 20 to 2022 March 7 and during morning astronomical twilight sky from 2019 Sep 21 to 2022 Sep 29. More than 46,000 exposures were taken in evening and morning astronomical twilight within 31 to 66 degrees from the Sun with an r-band limiting magnitude between 18.1 and 20.9. The twilight pointings show a slight seasonal dependence in limiting magnitude and ability to point closer towards the Sun, with limiting magnitude slightly improving during summer. In total, the one Aylo, (594913) 'Ayl\'o'chaxnim, and 4 Atiras, 2020 OV1, 2021 BS1, 2021 PB2, and 2021 VR3, were discovered in evening and morning twilight observations. Additional twilight survey discoveries also include 6 long-period comets: C/2020 T2, C/2020 V2, C/2021 D2, C/2021 E3, C/2022 E3, and C/2022 P3, and two short-period comets: P/2021 N1 and P/2022 P2 using deep learning comet detection pipelines. The P48/ZTF twilight survey also recovered 11 known Atiras, one Aylo, three short-period comes, two long-period comets, and one interstellar object. Lastly, the Vera Rubin Observatory will conduct a twilight survey starting in its first year of operations and will cover the sky within 45 degrees of the Sun. Twilight surveys such as those by ZTF and future surveys will provide opportunities for discovering asteroids inside the orbits of Earth and Venus., Comment: 26 pages, 13 figures, 4 tables, accepted for publication in Icarus

In this chapter, we will discuss the use of Machine Learning methods for the identification and localization of cometary activity for Solar System objects in ground and in space-based wide-field all-sky surveys. We will begin the chapter by discussing the challenges of identifying known and unknown active, extended Solar System objects in the presence of stellar-type sources and the application of classical pre-ML identification techniques and their limitations. We will then transition to the discussion of implementing ML techniques to address the challenge of extended object identification. We will finish with prospective future methods and the application to future surveys such as the Vera C. Rubin Observatory., Comment: 25 pages, 9 figures, accepted chapter in Machine Learning for Small Bodies in the Solar System, Valerio Carruba, Evgeny Smirnov, and Dagmara Oszkiewicz, Elsevier, 2024, p. 209-227

Visual bird's eye view (BEV) perception, due to its excellent perceptual capabilities, is progressively replacing costly LiDAR-based perception systems, especially in the realm of urban intelligent driving. However, this type of perception still relies on LiDAR data to construct ground truth databases, a process that is both cumbersome and time-consuming. Moreover, most massproduced autonomous driving systems are only equipped with surround camera sensors and lack LiDAR data for precise annotation. To tackle this challenge, we propose a fine-tuning method for BEV perception network based on visual 2D semantic perception, aimed at enhancing the model's generalization capabilities in new scene data. Considering the maturity and development of 2D perception technologies, our method significantly reduces the dependency on high-cost BEV ground truths and shows promising industrial application prospects. Extensive experiments and comparative analyses conducted on the nuScenes and Waymo public datasets demonstrate the effectiveness of our proposed method.

Social interactions form the foundation of human societies. Artificial intelligence has made significant progress in certain areas, but enabling machines to seamlessly understand social interactions remains an open challenge. It is important to address this gap by endowing machines with social capabilities. We identify three key capabilities needed for effective social understanding: 1) understanding multimodal social cues, 2) understanding multi-party dynamics, and 3) understanding beliefs. Building upon these foundations, we classify and review existing machine learning works on social understanding from the perspectives of verbal, non-verbal, and multimodal social cues. The verbal branch focuses on understanding linguistic signals such as speaker intent, dialogue sentiment, and commonsense reasoning. The non-verbal branch addresses techniques for perceiving social meaning from visual behaviors such as body gestures, gaze patterns, and facial expressions. The multimodal branch covers approaches that integrate verbal and non-verbal multimodal cues to holistically interpret social interactions such as recognizing emotions, conversational dynamics, and social situations. By reviewing the scope and limitations of current approaches and benchmarks, we aim to clarify the development trajectory and illuminate the path towards more comprehensive intelligence for social understanding. We hope this survey will spur further research interest and insights into this area.

The aim of our study is to characterise the spin states of the members of the Athor and Zita collisional families and test whether these members have a spin distribution consistent with a common origin from the break up of their respective family parent asteroids. Our method is based on the asteroid family evolution, which indicates that there should be a statistical predominance of retrograde-rotating asteroids on the inward side of family's V-shape, and prograde-rotating asteroids on the outward side. We used photometric data from our campaign and the literature in order to reveal the spin states of the asteroids belonging to these families. We combined dense and sparse photometric data in order to construct lightcurves; we performed the lightcurve inversion method to estimate the sidereal period, spin axis and convex shape of several family members. We obtained 34 new asteroid models for Athor family members and 17 for Zita family members. Along with the literature and revised models, the Athor family contains 60% of retrograde asteroids on the inward side and, 76% of prograde asteroids on the outward side. We also found that the Zita family exhibits 80% of retrograde asteroids on the inward side and an equal amount of prograde and retrograde rotators on the outward side. However, when we applied Kernel density estimation, we also found a clear peak for prograde asteroids on the outward side, as expected from the theory. The spin states of these asteroids validate the existence of both families, with the Athor family exhibiting a stronger signature for the presence of retrograde-rotating and prograde-rotating asteroids on the inner and outer side of the family, respectively. Our work provides an independent confirmation and characterisation of these very old families, whose presence and characteristics offer constraints for theories and models of the Solar System's evolution.

Federated learning (FL) has emerged as a promising paradigm for fine-tuning foundation models using distributed data in a privacy-preserving manner. Under limited computational resources, clients often find it more practical to fine-tune a selected subset of layers, rather than the entire model, based on their task-specific data. In this study, we provide a thorough theoretical exploration of selective layer fine-tuning in FL, emphasizing a flexible approach that allows the clients to adjust their selected layers according to their local data and resources. We theoretically demonstrate that the layer selection strategy has a significant impact on model convergence in two critical aspects: the importance of selected layers and the heterogeneous choices across clients. Drawing from these insights, we further propose a strategic layer selection method that utilizes local gradients and regulates layer selections across clients. The extensive experiments on both image and text datasets demonstrate the effectiveness of the proposed strategy compared with several baselines, highlighting its advances in identifying critical layers that adapt to the client heterogeneity and training dynamics in FL.

We introduce LLaVA-MoD, a novel framework designed to enable the efficient training of small-scale Multimodal Language Models (s-MLLM) by distilling knowledge from large-scale MLLM (l-MLLM). Our approach tackles two fundamental challenges in MLLM distillation. First, we optimize the network structure of s-MLLM by integrating a sparse Mixture of Experts (MoE) architecture into the language model, striking a balance between computational efficiency and model expressiveness. Second, we propose a progressive knowledge transfer strategy to ensure comprehensive knowledge migration. This strategy begins with mimic distillation, where we minimize the Kullback-Leibler (KL) divergence between output distributions to enable the student model to emulate the teacher network's understanding. Following this, we introduce preference distillation via Direct Preference Optimization (DPO), where the key lies in treating l-MLLM as the reference model. During this phase, the s-MLLM's ability to discriminate between superior and inferior examples is significantly enhanced beyond l-MLLM, leading to a better student that surpasses its teacher, particularly in hallucination benchmarks. Extensive experiments demonstrate that LLaVA-MoD outperforms existing models across various multimodal benchmarks while maintaining a minimal number of activated parameters and low computational costs. Remarkably, LLaVA-MoD, with only 2B activated parameters, surpasses Qwen-VL-Chat-7B by an average of 8.8% across benchmarks, using merely 0.3% of the training data and 23% trainable parameters. These results underscore LLaVA-MoD's ability to effectively distill comprehensive knowledge from its teacher model, paving the way for the development of more efficient MLLMs. The code will be available on: https://github.com/shufangxun/LLaVA-MoD.

Scoring rules are aimed at evaluation of the quality of predictions, but can also be used for estimation of parameters in statistical models. We propose estimating parameters of multivariate spatial models by maximising the average leave-one-out cross-validation score. This method, LOOS, thus optimises predictions instead of maximising the likelihood. The method allows for fast computations for Gaussian models with sparse precision matrices, such as spatial Markov models. It also makes it possible to tailor the estimator's robustness to outliers and their sensitivity to spatial variations of uncertainty through the choice of the scoring rule which is used in the maximisation. The effects of the choice of scoring rule which is used in LOOS are studied by simulation in terms of computation time, statistical efficiency, and robustness. Various popular scoring rules and a new scoring rule, the root score, are compared to maximum likelihood estimation. The results confirmed that for spatial Markov models the computation time for LOOS was much smaller than for maximum likelihood estimation. Furthermore, the standard deviations of parameter estimates were smaller for maximum likelihood estimation, although the differences often were small. The simulations also confirmed that the usage of a robust scoring rule results in robust LOOS estimates and that the robustness provides better predictive quality for spatial data with outliers. Finally, the new inference method was applied to ERA5 temperature reanalysis data for the contiguous United States and the average July temperature for the years 1940 to 2023, and this showed that the LOOS estimator provided parameter estimates that were more than a hundred times faster to compute compared to maximum-likelihood estimation, and resulted in a model with better predictive performance.

Photostriction is a phenomenon that can potentially improve the precision of light-driven actuation, the sensitivity of photodetection, and the efficiency of optical energy harvesting. However, known materials with significant photostriction are limited, and effective guidelines to discover new photostrictive materials are lacking. In this study, we perform a high-throughput computational search for new photostrictive materials based on simple thermodynamic descriptors, namely the band gap pressure and stress coefficients. Using constrained density functional theory simulations, we establish that these descriptors can accurately predict intrinsic photostriction in a wide range of materials. Subsequently, we screen over 4770 stable semiconductors with a band gap below 2 eV from the Materials Project database to search for strongly photostrictive materials. This search identifies PtS$_2$ and Te$_2$I as the most promising ones, with photostriction exceeding 10$^{-4}$ with a moderate photocarrier concentration of 10$^{18}$ cm$^{-3}$. Furthermore, we provide a detailed analysis of factors contributing to strong photostriction, including bulk moduli and band-edge orbital interactions. Our results provide physical insights into photostriction of materials and demonstrate the effectiveness of using simple descriptors in high-throughput searches for new functional materials.

This study introduces the 4D Risk Occupancy within a vehicle-road-cloud architecture, integrating the road surface spatial, risk, and temporal dimensions, and endowing the algorithm with beyond-line-of-sight, all-angles, and efficient abilities. The algorithm simplifies risk modeling by focusing on directly observable information and key factors, drawing on the concept of Occupancy Grid Maps (OGM), and incorporating temporal prediction to effectively map current and future risk occupancy. Compared to conventional driving risk fields and grid occupancy maps, this algorithm can map global risks more efficiently, simply, and reliably. It can integrate future risk information, adapting to dynamic traffic environments. The 4D Risk Occupancy also unifies the expression of BEV detection and lane line detection results, enhancing the intuitiveness and unity of environmental perception. Using DAIR-V2X data, this paper validates the 4D Risk Occupancy algorithm and develops a local path planning model based on it. Qualitative experiments under various road conditions demonstrate the practicality and robustness of this local path planning model. Quantitative analysis shows that the path planning based on risk occupation significantly improves trajectory planning performance, increasing safety redundancy by 12.5% and reducing average deceleration by 5.41% at an initial braking speed of 8 m/s, thereby improving safety and comfort. This work provides a new global perception method and local path planning method through Vehicle-Road-Cloud architecture, offering a new perceptual paradigm for achieving safer and more efficient autonomous driving., Comment: 13 pages,9 figures

In econometrics, the Efficient Market Hypothesis posits that asset prices reflect all available information in the market. Several empirical investigations show that market efficiency drops when it undergoes extreme events. Many models for multivariate extremes focus on positive dependence, making them unsuitable for studying extremal dependence in financial markets where data often exhibit both positive and negative extremal dependence. To this end, we construct regular variation models on the entirety of $\mathbb{R}^d$ and develop a bivariate measure for asymmetry in the strength of extremal dependence between adjacent orthants. Our directional tail dependence (DTD) measure allows us to define the Efficient Tail Hypothesis (ETH) -- an analogue of the Efficient Market Hypothesis -- for the extremal behaviour of the market. Asymptotic results for estimators of DTD are described, and we discuss testing of the ETH via permutation-based methods and present novel tools for visualization. Empirical study of China's futures market leads to a rejection of the ETH and we identify potential profitable investment opportunities. To promote the research of microstructure in China's derivatives market, we open-source our high-frequency data, which are being collected continuously from multiple derivative exchanges., Comment: 56 pages, 9 figures, 3 tables

Federated learning (FL) becomes vulnerable to Byzantine attacks where some of participators tend to damage the utility or discourage the convergence of the learned model via sending their malicious model updates. Previous works propose to apply robust rules to aggregate updates from participators against different types of Byzantine attacks, while at the same time, attackers can further design advanced Byzantine attack algorithms targeting specific aggregation rule when it is known. In practice, FL systems can involve a black-box server that makes the adopted aggregation rule inaccessible to participants, which can naturally defend or weaken some Byzantine attacks. In this paper, we provide an in-depth understanding on the Byzantine robustness of the FL system with a black-box server. Our investigation demonstrates the improved Byzantine robustness of a black-box server employing a dynamic defense strategy. We provide both empirical evidence and theoretical analysis to reveal that the black-box server can mitigate the worst-case attack impact from a maximum level to an expectation level, which is attributed to the inherent inaccessibility and randomness offered by a black-box server.The source code is available at https://github.com/alibaba/FederatedScope/tree/Byzantine_attack_defense to promote further research in the community., Comment: We have released code on https://github.com/alibaba/FederatedScope/tree/Byzantine_attack_defense

We extend the parity doublet model for hadronic matter and study the possible presence of quark matter inside the cores of neutron stars with the Nambu-Jona-Lasinio (NJL) model. Considering the uncertainties of the QCD phase diagram and the location of the critical endpoint, we aim to explore the competition between the chiral phase transition and the deconfinement phase transition systematically, regulated by the vacuum pressure $-B$ in the NJL model. Employing a Maxwell construction, a sharp first-order deconfinement phase transition is implemented combining the parity doublet model for the hadronic phase and the NJL model for the high-energy quark phase. The position of the chiral phase transition is obtained from the NJL model self-consistently. We find stable neutron stars with a quark core within a specific parameter space that satisfies current astronomical observations. The observations suggest a relatively large chiral invariant mass $m_0=600$ MeV in the parity doublet model and a larger split between the chiral and deconfinement phase transitions while assuming the first-order deconfinement phase transition. The maximum mass of the hybrid star that we obtain is $\sim 2.2 M_{\odot}$., Comment: 10pages,7 figures

The unique diagnosis and treatment techniques and remarkable clinical efficacy of traditional Chinese medicine (TCM) make it play an important role in the field of elderly care and healthcare, especially in the rehabilitation of some common chronic diseases of the elderly. Therefore, building a TCM chatbot for healthcare application will help users obtain consultation services in a direct and natural way. However, concepts such as acupuncture points (acupoints) and meridians involved in TCM always appear in the consultation, which cannot be displayed intuitively. To this end, we develop a \textbf{h}ealthcare chat\textbf{bot} (HBot) based on a human body model in 3D and knowledge graph, which provides conversational services such as knowledge Q\&A, prescription recommendation, moxibustion therapy recommendation, and acupoint search. When specific acupoints are involved in the conversations between user and HBot, the 3D body will jump to the corresponding acupoints and highlight them. Moreover, Hbot can also be used in training scenarios to accelerate the teaching process of TCM by intuitively displaying acupuncture points and knowledge cards. The demonstration video is available at https://www.youtube.com/watch?v=UhQhutSKkTU . Our code and dataset are publicly available at Gitee: https://gitee.com/plabrolin/interactive-3d-acup.git, Comment: System Demonstration

Recent advances in large language models (LLMs) have opened new avenues for applying multi-agent systems in very large-scale simulations. However, there remain several challenges when conducting multi-agent simulations with existing platforms, such as limited scalability and low efficiency, unsatisfied agent diversity, and effort-intensive management processes. To address these challenges, we develop several new features and components for AgentScope, a user-friendly multi-agent platform, enhancing its convenience and flexibility for supporting very large-scale multi-agent simulations. Specifically, we propose an actor-based distributed mechanism as the underlying technological infrastructure towards great scalability and high efficiency, and provide flexible environment support for simulating various real-world scenarios, which enables parallel execution of multiple agents, centralized workflow orchestration, and both inter-agent and agent-environment interactions among agents. Moreover, we integrate an easy-to-use configurable tool and an automatic background generation pipeline in AgentScope, simplifying the process of creating agents with diverse yet detailed background settings. Last but not least, we provide a web-based interface for conveniently monitoring and managing a large number of agents that might deploy across multiple devices. We conduct a comprehensive simulation to demonstrate the effectiveness of the proposed enhancements in AgentScope, and provide detailed observations and discussions to highlight the great potential of applying multi-agent systems in large-scale simulations. The source code is released on GitHub at https://github.com/modelscope/agentscope to inspire further research and development in large-scale multi-agent simulations., Comment: We have released code on https://github.com/modelscope/agentscope

We initiate a formal investigation into the design and analysis of LLM-based algorithms, i.e. algorithms that contain one or multiple calls of large language models (LLMs) as sub-routines and critically rely on the capabilities of LLMs. While LLM-based algorithms, ranging from basic LLM calls with prompt engineering to complicated LLM-powered agent systems and compound AI systems, have achieved remarkable empirical success, the design and optimization of them have mostly relied on heuristics and trial-and-errors, which is largely due to a lack of formal and analytical study for these algorithms. To fill this gap, we start by identifying the computational-graph representation of LLM-based algorithms, the design principle of task decomposition, and some key abstractions, which then facilitate our formal analysis for the accuracy and efficiency of LLM-based algorithms, despite the black-box nature of LLMs. Through extensive analytical and empirical investigation in a series of case studies, we demonstrate that the proposed framework is broadly applicable to a wide range of scenarios and diverse patterns of LLM-based algorithms, such as parallel, hierarchical and recursive task decomposition. Our proposed framework holds promise for advancing LLM-based algorithms, by revealing the reasons behind curious empirical phenomena, guiding the choices of hyperparameters, predicting the empirical performance of algorithms, and inspiring new algorithm design. To promote further study of LLM-based algorithms, we release our source code at https://github.com/modelscope/agentscope/tree/main/examples/paper_llm_based_algorithm.

The emergence of large-scale multi-modal generative models has drastically advanced artificial intelligence, introducing unprecedented levels of performance and functionality. However, optimizing these models remains challenging due to historically isolated paths of model-centric and data-centric developments, leading to suboptimal outcomes and inefficient resource utilization. In response, we present a novel sandbox suite tailored for integrated data-model co-development. This sandbox provides a comprehensive experimental platform, enabling rapid iteration and insight-driven refinement of both data and models. Our proposed "Probe-Analyze-Refine" workflow, validated through applications on state-of-the-art LLaVA-like and DiT based models, yields significant performance boosts, such as topping the VBench leaderboard. We also uncover fruitful insights gleaned from exhaustive benchmarks, shedding light on the critical interplay between data quality, diversity, and model behavior. With the hope of fostering deeper understanding and future progress in multi-modal data and generative modeling, our codes, datasets, and models are maintained and accessible at https://github.com/modelscope/data-juicer/blob/main/docs/Sandbox.md., Comment: 26 pages, 9 figures, 5 tables

Direct Preference Optimization (DPO) has emerged as a compelling approach for training Large Language Models (LLMs) to adhere to human preferences. However, the performance of DPO is sensitive to the fine-tuning of its trade-off parameter $\beta$, as well as to the quality of the preference data. We analyze the impact of $\beta$ and data quality on DPO, uncovering that optimal $\beta$ values vary with the informativeness of pairwise data. Addressing the limitations of static $\beta$ values, we introduce a novel framework that dynamically calibrates $\beta$ at the batch level, informed by data quality considerations. Additionally, our method incorporates $\beta$-guided data filtering to safeguard against the influence of outliers. Through empirical evaluation, we demonstrate that our dynamic $\beta$ adjustment technique significantly improves DPO's performance across a range of models and datasets, offering a more robust and adaptable training paradigm for aligning LLMs with human feedback. The code is available at \url{https://github.com/junkangwu/beta-DPO}.

The rapid development of large language models (LLMs) has been witnessed in recent years. Based on the powerful LLMs, multi-modal LLMs (MLLMs) extend the modality from text to a broader spectrum of domains, attracting widespread attention due to the broader range of application scenarios. As LLMs and MLLMs rely on vast amounts of model parameters and data to achieve emergent capabilities, the importance of data is receiving increasingly widespread attention and recognition. Tracing and analyzing recent data-oriented works for MLLMs, we find that the development of models and data is not two separate paths but rather interconnected. On the one hand, vaster and higher-quality data contribute to better performance of MLLMs; on the other hand, MLLMs can facilitate the development of data. The co-development of multi-modal data and MLLMs requires a clear view of 1) at which development stages of MLLMs specific data-centric approaches can be employed to enhance certain MLLM capabilities, and 2) how MLLMs, utilizing those capabilities, can contribute to multi-modal data in specific roles. To promote the data-model co-development for MLLM community, we systematically review existing works related to MLLMs from the data-model co-development perspective. A regularly maintained project associated with this survey is accessible at https://github.com/modelscope/data-juicer/blob/main/docs/awesome_llm_data.md., Comment: Ongoing work. 21 pages. Related materials are continually maintained and available at https://github.com/modelscope/data-juicer/blob/main/docs/awesome_llm_data.md

This study addresses the challenge of noise in training datasets for Direct Preference Optimization (DPO), a method for aligning Large Language Models (LLMs) with human preferences. We categorize noise into pointwise noise, which includes low-quality data points, and pairwise noise, which encompasses erroneous data pair associations that affect preference rankings. Utilizing Distributionally Robust Optimization (DRO), we enhance DPO's resilience to these types of noise. Our theoretical insights reveal that DPO inherently embeds DRO principles, conferring robustness to pointwise noise, with the regularization coefficient $\beta$ playing a critical role in its noise resistance. Extending this framework, we introduce Distributionally Robustifying DPO (Dr. DPO), which integrates pairwise robustness by optimizing against worst-case pairwise scenarios. The novel hyperparameter $\beta'$ in Dr. DPO allows for fine-tuned control over data pair reliability, providing a strategic balance between exploration and exploitation in noisy training environments. Empirical evaluations demonstrate that Dr. DPO substantially improves the quality of generated text and response accuracy in preference datasets, showcasing enhanced performance in both noisy and noise-free settings. The code is available at https://github.com/junkangwu/Dr_DPO.

Data synthesis is a promising solution to share data for various downstream analytic tasks without exposing raw data. However, without a theoretical privacy guarantee, a synthetic dataset would still leak some sensitive information. Differential privacy is thus widely adopted to safeguard data synthesis by strictly limiting the released information. This technique is advantageous yet presents significant challenges in the vertical federated setting, where data attributes are distributed among different data parties. The main challenge lies in maintaining privacy while efficiently and precisely reconstructing the correlation among cross-party attributes. In this paper, we propose a novel algorithm called VertiMRF, designed explicitly for generating synthetic data in the vertical setting and providing differential privacy protection for all information shared from data parties. We introduce techniques based on the Flajolet-Martin sketch (or frequency oracle) for encoding local data satisfying differential privacy and estimating cross-party marginals. We provide theoretical privacy and utility proof for encoding in this multi-attribute data. Collecting the locally generated private Markov Random Field (MRF) and the sketches, a central server can reconstruct a global MRF, maintaining the most useful information. Additionally, we introduce two techniques tailored for datasets with large attribute domain sizes, namely dimension reduction and consistency enforcement. These two techniques allow flexible and inconsistent binning strategies of local private MRF and the data sketching module, which can preserve information to the greatest extent. We conduct extensive experiments on four real-world datasets to evaluate the effectiveness of VertiMRF. End-to-end comparisons demonstrate the superiority of VertiMRF, and ablation studies validate the effectiveness of each component.

Large language models (LLMs) show amazing performance on many domain-specific tasks after fine-tuning with some appropriate data. However, many domain-specific data are privately distributed across multiple owners. Thus, this dilemma raises the interest in how to perform LLM fine-tuning in federated learning (FL). However, confronted with limited computation and communication capacities, FL clients struggle to fine-tune an LLM effectively. To this end, we introduce FedBiOT, a resource-efficient LLM fine-tuning approach to FL. Specifically, our method involves the server generating a compressed LLM and aligning its performance with the full model. Subsequently, the clients fine-tune a lightweight yet important part of the compressed model, referred to as an adapter. Notice that as the server has no access to the private data owned by the clients, the data used for alignment by the server has a different distribution from the one used for fine-tuning by clients. We formulate the problem into a bi-level optimization problem to minimize the negative effect of data discrepancy and derive the updating rules for the server and clients. We conduct extensive experiments on LLaMA-2, empirically showing that the adapter has exceptional performance when reintegrated into the global LLM. The results also indicate that the proposed FedBiOT significantly reduces resource consumption compared to existing benchmarks, all while achieving comparable performance levels., Comment: KDD 2024

Spurious bias, a tendency to use spurious correlations between non-essential input attributes and target variables for predictions, has revealed a severe robustness pitfall in deep learning models trained on single modality data. Multimodal Large Language Models (MLLMs), which integrate both vision and language models, have demonstrated strong capability in joint vision-language understanding. However, whether spurious biases are prevalent in MLLMs remains under-explored. We mitigate this gap by analyzing the spurious biases in a multimodal setting, uncovering the specific test data patterns that can manifest this problem when biases in the vision model cascade into the alignment between visual and text tokens in MLLMs. To better understand this problem, we introduce MM-SpuBench, a comprehensive visual question-answering (VQA) benchmark designed to evaluate MLLMs' reliance on nine distinct categories of spurious correlations from five open-source image datasets. The VQA dataset is built from human-understandable concept information (attributes). Leveraging this benchmark, we conduct a thorough evaluation of current state-of-the-art MLLMs. Our findings illuminate the persistence of the reliance on spurious correlations from these models and underscore the urge for new methodologies to mitigate spurious biases. To support the MLLM robustness research, we release our VQA benchmark at https://huggingface.co/datasets/mmbench/MM-SpuBench.

Recently, Multimodal Large Language Models (MLLMs) have shown great promise in language-guided perceptual tasks such as recognition, segmentation, and object detection. However, their effectiveness in addressing visual cognition problems that require high-level reasoning is not well-established. One such challenge is abstract visual reasoning (AVR) -- the cognitive ability to discern relationships among patterns in a set of images and extrapolate to predict subsequent patterns. This skill is crucial during the early neurodevelopmental stages of children. Inspired by the AVR tasks in Raven's Progressive Matrices (RPM) and Wechsler Intelligence Scale for Children (WISC), we propose a new dataset MaRs-VQA and a new benchmark VCog-Bench containing three datasets to evaluate the zero-shot AVR capability of MLLMs and compare their performance with existing human intelligent investigation. Our comparative experiments with different open-source and closed-source MLLMs on the VCog-Bench revealed a gap between MLLMs and human intelligence, highlighting the visual cognitive limitations of current MLLMs. We believe that the public release of VCog-Bench, consisting of MaRs-VQA, and the inference pipeline will drive progress toward the next generation of MLLMs with human-like visual cognition abilities., Comment: 14 pages, 4 figures, the appendix will be updated soon

Photo-doped Mott insulators can exhibit novel photocarrier transport and relaxation dynamics and non-equilibrium phases. However, time-resolved real-space imaging of these processes are still lacking. Here, we use scanning ultrafast electron microscopy (SUEM) to directly visualize the spatial-temporal evolution of photoexcited species in a spin-orbit assisted Mott insulator {\alpha}-RuCl3. At low optical fluences, we observe extremely long hot photocarrier transport time over one nanosecond, almost an order of magnitude longer than any known values in conventional semiconductors. At higher optical fluences, we observe nonlinear features suggesting a photo-induced insulator-to-metal transition, which is unusual in a large-gap Mott insulator. Our results demonstrate the rich physics in a photo-doped Mott insulator that can be extracted from spatial-temporal imaging and showcase the capability of SUEM to sensitively probe photoexcitations in strongly correlated electron systems., Comment: Comments are welcome. Please email feedback to bliao@ucsb.edu

Cardinality estimation (CardEst) is essential for optimizing query execution plans. Recent ML-based CardEst methods achieve high accuracy but face deployment challenges due to high preparation costs and lack of transferability across databases. In this paper, we propose PRICE, a PRetrained multI-table CardEst model, which addresses these limitations. PRICE takes low-level but transferable features w.r.t. data distributions and query information and elegantly applies self-attention models to learn meta-knowledge to compute cardinality in any database. It is generally applicable to any unseen new database to attain high estimation accuracy, while its preparation cost is as little as the basic one-dimensional histogram-based CardEst methods. Moreover, PRICE can be finetuned to further enhance its performance on any specific database. We pretrained PRICE using 30 diverse datasets, completing the process in about 5 hours with a resulting model size of only about 40MB. Evaluations show that PRICE consistently outperforms existing methods, achieving the highest estimation accuracy on several unseen databases and generating faster execution plans with lower overhead. After finetuning with a small volume of databasespecific queries, PRICE could even find plans very close to the optimal ones. Meanwhile, PRICE is generally applicable to different settings such as data updates, data scaling, and query workload shifts. We have made all of our data and codes publicly available at https://github.com/StCarmen/PRICE.

In this work, we present Xwin-LM, a comprehensive suite of alignment methodologies for large language models (LLMs). This suite encompasses several key techniques, including supervised finetuning (SFT), reward modeling (RM), rejection sampling finetuning (RS), and direct preference optimization (DPO). The key components are as follows: (1) Xwin-LM-SFT, models initially finetuned with high-quality instruction data; (2) Xwin-Pair, a large-scale, multi-turn preference dataset meticulously annotated using GPT-4; (3) Xwin-RM, reward models trained on Xwin-Pair, developed at scales of 7B, 13B, and 70B parameters; (4) Xwin-Set, a multiwise preference dataset in which each prompt is linked to 64 unique responses generated by Xwin-LM-SFT and scored by Xwin-RM; (5) Xwin-LM-RS, models finetuned with the highest-scoring responses from Xwin-Set; (6) Xwin-LM-DPO, models further optimized on Xwin-Set using the DPO algorithm. Our evaluations on AlpacaEval and MT-bench demonstrate consistent and significant improvements across the pipeline, demonstrating the strength and scalability of Xwin-LM. The repository https://github.com/Xwin-LM/Xwin-LM will be continually updated to foster community research.

Large language models exhibit exceptional generalization capabilities, primarily attributed to the utilization of diversely sourced data. However, conventional practices in integrating this diverse data heavily rely on heuristic schemes, lacking theoretical guidance. This research tackles these limitations by investigating strategies based on low-cost proxies for data mixtures, with the aim of streamlining data curation to enhance training efficiency. Specifically, we propose a unified scaling law, termed $\textbf{BiMix}$, which accurately models the bivariate scaling behaviors of both data quantity and mixing proportions. We conduct systematic experiments and provide empirical evidence for the predictive power and fundamental principles of $\textbf{BiMix}$. Notably, our findings reveal that entropy-driven training-free data mixtures can achieve comparable or even better performance than more resource-intensive methods. We hope that our quantitative insights can shed light on further judicious research and development in cost-effective language modeling., Comment: typos corrected

Manipulating thermal properties in materials has been of fundamental importance for advancing innovative technologies. Heat carriers such as phonons are impeded by breaking crystal symmetry or periodicity. Notable methods of impeding the phonon propagation include varying the density of defects, interfaces, and nanostructures, as well as changing composition. However, a robust link between the individual nanoscale defect structures, phonon states, and macroscopic thermal conductivity is lacking. Here we reveal from nanoscale structure-phonon mechanisms on how the grain boundary (GB) tilt and twist angles fundamentally drive the changes in atom rearrangements, exotic vibrational states, and finally macroscopic heat transport at different bicrystal strontium titanate GBs using emerging atomic resolution vibrational spectroscopy. The 10 deg and 22 deg tilt GBs exhibit reduced phonon populations by 54% and 16% compared to the bulk value, respectively, consistent with measured thermal conductivities. A tiny twist angle further introduces a fine and local tunning of thermal conductivity by introducing twist induced defects periodically embedded with the tilt induced GB defects. Our results demonstrate that varying the tilt angle coarsely modifies the phonon population along entire GB while varying the twist angle incurs a finer adjustment at periodic locations on the GB. Our study offers a systematic approach to understanding and manipulating cross GB thermal transport of arbitrary GBs predictably and precisely.

Designing effective data manipulation methods is a long standing problem in data lakes. Traditional methods, which rely on rules or machine learning models, require extensive human efforts on training data collection and tuning models. Recent methods apply Large Language Models (LLMs) to resolve multiple data manipulation tasks. They exhibit bright benefits in terms of performance but still require customized designs to fit each specific task. This is very costly and can not catch up with the requirements of big data lake platforms. In this paper, inspired by the cross-task generality of LLMs on NLP tasks, we pave the first step to design an automatic and general solution to tackle with data manipulation tasks. We propose UniDM, a unified framework which establishes a new paradigm to process data manipulation tasks using LLMs. UniDM formalizes a number of data manipulation tasks in a unified form and abstracts three main general steps to solve each task. We develop an automatic context retrieval to allow the LLMs to retrieve data from data lakes, potentially containing evidence and factual information. For each step, we design effective prompts to guide LLMs to produce high quality results. By our comprehensive evaluation on a variety of benchmarks, our UniDM exhibits great generality and state-of-the-art performance on a wide variety of data manipulation tasks., Comment: MLSys24

Semiconductor heterojunctions have gained significant attention for efficient optoelectronic devices owing to their unique interfaces and synergistic effects. Interaction between charge carriers with the heterojunction plays a crucial role in determining device performance, while its spatial-temporal mapping remains lacking. In this study, we employ scanning ultrafast electron microscopy (SUEM), an emerging technique that combines high spatial-temporal resolution and surface sensitivity, to investigate photocarrier dynamics across a Si/Ge heterojunction. Charge dynamics are selectively examined across the junction and compared to far bulk areas, through which the impact of the built-in potential, band offsets, and surface effects is directly visualized. In particular, we find that the heterojunction drastically modifies the hot photocarrier diffusivities by up to 300%. These findings are further elucidated with insights from the band structure and surface potential measured by complementary techniques. This work demonstrates the tremendous effect of heterointerfaces on charge dynamics and showcases the potential of SUEM in characterizing realistic devices.

Magnetocaloric materials, which exploit reversible temperature changes induced by magnetic field variations, are promising for advancing energy-efficient cooling technologies. The potential integration of two-dimensional materials into magnetocaloric systems represents an emerging opportunity to enhance the magnetocaloric cooling efficiency. In this study, we use atomistic spin dynamics simulations based on first-principles parameters to systematically evaluate how magnetocaloric properties transition from three-dimensional (3D) to two-dimensional (2D) ferromagnetic materials. We find that 2D features such as reduced Curie temperature, sharper magnetic transition, and higher magnetic susceptibility are beneficial for magnetocaloric applications, while the relatively higher lattice heat capacity in 2D can compromise achievable adiabatic temperature changes. We further propose GdSi$_2$ as a promising 2D magnetocaloric material near hydrogen liquefaction temperature. Our analysis offers valuable theoretical insights into the magnetocaloric effect in 2D ferromagnets and demonstrates that 2D ferromagnets hold promise for cooling and thermal management applications in compact and miniaturized nanodevices.

Two-dimensional (2D) materials have shown great potential in applications as transistors, where thermal dissipation becomes crucial because of the increasing energy density. Although thermal conductivity of 2D materials has been extensively studied, interactions between nonequilibrium electrons and phonons, which can be strong when high electric fields and heat current coexist, are not considered. In this work, we systematically study the electron drag effect, where nonequilibrium electrons impart momenta to phonons and influence the thermal conductivity, in 2D semiconductors using ab initio simulations. We find that, at room temperature, electron drag can significantly increase thermal conductivity by decreasing phonon-electron scattering in 2D semiconductors, while its impact in three-dimensional (3D) semiconductors is negligible. We attribute this difference to the large electron-phonon scattering phase space and higher contribution to thermal conductivity by drag-active phonons. Our work elucidates the fundamental physics underlying coupled electron-phonon transport in materials of various dimensionalities., Comment: 16 pages, 5 figures

Despite the success of large language models (LLMs) in natural language generation, much evidence shows that LLMs may produce incorrect or nonsensical text. This limitation highlights the importance of discerning when to trust LLMs, especially in safety-critical domains. Existing methods often express reliability by confidence level, however, their effectiveness is limited by the lack of objective guidance. To address this, we propose CONfidence-Quality-ORDer-preserving alignment approach (CONQORD), which leverages reinforcement learning guided by a tailored dual-component reward function. This function integrates quality reward and order-preserving alignment reward functions. Specifically, the order-preserving reward incentivizes the model to verbalize greater confidence for responses of higher quality to align the order of confidence and quality. Experiments demonstrate that CONQORD significantly improves the alignment performance between confidence and response accuracy, without causing over-cautious. Furthermore, the aligned confidence provided by CONQORD informs when to trust LLMs, and acts as a determinant for initiating the retrieval process of external knowledge. Aligning confidence with response quality ensures more transparent and reliable responses, providing better trustworthiness., Comment: Accepted by ACL 2024. Code: https://github.com/TaoShuchang/CONQORD

Text animation serves as an expressive medium, transforming static communication into dynamic experiences by infusing words with motion to evoke emotions, emphasize meanings, and construct compelling narratives. Crafting animations that are semantically aware poses significant challenges, demanding expertise in graphic design and animation. We present an automated text animation scheme, termed "Dynamic Typography", which combines two challenging tasks. It deforms letters to convey semantic meaning and infuses them with vibrant movements based on user prompts. Our technique harnesses vector graphics representations and an end-to-end optimization-based framework. This framework employs neural displacement fields to convert letters into base shapes and applies per-frame motion, encouraging coherence with the intended textual concept. Shape preservation techniques and perceptual loss regularization are employed to maintain legibility and structural integrity throughout the animation process. We demonstrate the generalizability of our approach across various text-to-video models and highlight the superiority of our end-to-end methodology over baseline methods, which might comprise separate tasks. Through quantitative and qualitative evaluations, we demonstrate the effectiveness of our framework in generating coherent text animations that faithfully interpret user prompts while maintaining readability. Our code is available at: https://animate-your-word.github.io/demo/., Comment: Our demo page is available at: https://animate-your-word.github.io/demo/

The environmental perception of autonomous vehicles in normal conditions have achieved considerable success in the past decade. However, various unfavourable conditions such as fog, low-light, and motion blur will degrade image quality and pose tremendous threats to the safety of autonomous driving. That is, when applied to degraded images, state-of-the-art visual models often suffer performance decline due to the feature content loss and artifact interference caused by statistical and structural properties disruption of captured images. To address this problem, this work proposes a novel Deep Channel Prior (DCP) for degraded visual recognition. Specifically, we observe that, in the deep representation space of pre-trained models, the channel correlations of degraded features with the same degradation type have uniform distribution even if they have different content and semantics, which can facilitate the mapping relationship learning between degraded and clear representations in high-sparsity feature space. Based on this, a novel plug-and-play Unsupervised Feature Enhancement Module (UFEM) is proposed to achieve unsupervised feature correction, where the multi-adversarial mechanism is introduced in the first stage of UFEM to achieve the latent content restoration and artifact removal in high-sparsity feature space. Then, the generated features are transferred to the second stage for global correlation modulation under the guidance of DCP to obtain high-quality and recognition-friendly features. Evaluations of three tasks and eight benchmark datasets demonstrate that our proposed method can comprehensively improve the performance of pre-trained models in real degradation conditions. The source code is available at https://github.com/liyuhang166/Deep_Channel_Prior, Comment: 14 pages, 14 figures, publised to TIV2024

Local differential privacy (LDP) is a strong privacy standard that has been adopted by popular software systems. The main idea is that each individual perturbs their own data locally, and only submits the resulting noisy version to a data aggregator. Although much effort has been devoted to computing various types of aggregates and building machine learning applications under LDP, research on fundamental perturbation mechanisms has not achieved significant improvement in recent years. Towards a more refined result utility, existing works mainly focus on improving the worst-case guarantee. However, this approach does not necessarily promise a better average performance given the fact that the data in practice obey a certain distribution, which is not known beforehand. In this paper, we propose the advanced adaptive additive (AAA) mechanism, which is a distribution-aware approach that addresses the average utility and tackles the classical mean estimation problem. AAA is carried out in a two-step approach: first, as the global data distribution is not available beforehand, the data aggregator selects a random subset of individuals to compute a (noisy) quantized data descriptor; then, the data aggregator collects data from the remaining individuals, which are perturbed in a distribution-aware fashion. The perturbation involved in the latter step is obtained by solving an optimization problem, which is formulated with the data descriptor obtained in the former step and the desired properties of task-determined utilities. We provide rigorous privacy proofs, utility analyses, and extensive experiments comparing AAA with state-of-the-art mechanisms. The evaluation results demonstrate that the AAA mechanism consistently outperforms existing solutions with a clear margin in terms of result utility, on a wide range of privacy constraints and real-world and synthetic datasets.

The estimation of regression parameters in spatially referenced data plays a crucial role across various scientific domains. A common approach involves employing an additive regression model to capture the relationship between observations and covariates, accounting for spatial variability not explained by the covariates through a Gaussian random field. While theoretical analyses of such models have predominantly focused on prediction and covariance parameter inference, recent attention has shifted towards understanding the theoretical properties of regression coefficient estimates, particularly in the context of spatial confounding. This article studies the effect of misspecified covariates, in particular when the misspecification changes the smoothness. We analyze the theoretical properties of the generalize least-square estimator under infill asymptotics, and show that the estimator can have counter-intuitive properties. In particular, the estimated regression coefficients can converge to zero as the number of observations increases, despite high correlations between observations and covariates. Perhaps even more surprising, the estimates can diverge to infinity under certain conditions. Through an application to temperature and precipitation data, we show that both behaviors can be observed for real data. Finally, we propose a simple fix to the problem by adding a smoothing step in the regression.

Continual learning (CL) aims to empower models to learn new tasks without forgetting previously acquired knowledge. Most prior works concentrate on the techniques of architectures, replay data, regularization, \etc. However, the category name of each class is largely neglected. Existing methods commonly utilize the one-hot labels and randomly initialize the classifier head. We argue that the scarce semantic information conveyed by the one-hot labels hampers the effective knowledge transfer across tasks. In this paper, we revisit the role of the classifier head within the CL paradigm and replace the classifier with semantic knowledge from pretrained language models (PLMs). Specifically, we use PLMs to generate semantic targets for each class, which are frozen and serve as supervision signals during training. Such targets fully consider the semantic correlation between all classes across tasks. Empirical studies show that our approach mitigates forgetting by alleviating representation drifting and facilitating knowledge transfer across tasks. The proposed method is simple to implement and can seamlessly be plugged into existing methods with negligible adjustments. Extensive experiments based on eleven mainstream baselines demonstrate the effectiveness and generalizability of our approach to various protocols. For example, under the class-incremental learning setting on ImageNet-100, our method significantly improves the Top-1 accuracy by 3.2\% to 6.1\% while reducing the forgetting rate by 2.6\% to 13.1\%., Comment: Accepted by CVPR 2024

We observe a high level of imbalance in the accuracy of different classes in the same old task for the first time. This intriguing phenomenon, discovered in replay-based Class Incremental Learning (CIL), highlights the imbalanced forgetting of learned classes, as their accuracy is similar before the occurrence of catastrophic forgetting. This discovery remains previously unidentified due to the reliance on average incremental accuracy as the measurement for CIL, which assumes that the accuracy of classes within the same task is similar. However, this assumption is invalid in the face of catastrophic forgetting. Further empirical studies indicate that this imbalanced forgetting is caused by conflicts in representation between semantically similar old and new classes. These conflicts are rooted in the data imbalance present in replay-based CIL methods. Building on these insights, we propose CLass-Aware Disentanglement (CLAD) to predict the old classes that are more likely to be forgotten and enhance their accuracy. Importantly, CLAD can be seamlessly integrated into existing CIL methods. Extensive experiments demonstrate that CLAD consistently improves current replay-based methods, resulting in performance gains of up to 2.56%., Comment: AAAI2024

For privacy and security concerns, the need to erase unwanted information from pre-trained vision models is becoming evident nowadays. In real-world scenarios, erasure requests originate at any time from both users and model owners. These requests usually form a sequence. Therefore, under such a setting, selective information is expected to be continuously removed from a pre-trained model while maintaining the rest. We define this problem as continual forgetting and identify two key challenges. (i) For unwanted knowledge, efficient and effective deleting is crucial. (ii) For remaining knowledge, the impact brought by the forgetting procedure should be minimal. To address them, we propose Group Sparse LoRA (GS-LoRA). Specifically, towards (i), we use LoRA modules to fine-tune the FFN layers in Transformer blocks for each forgetting task independently, and towards (ii), a simple group sparse regularization is adopted, enabling automatic selection of specific LoRA groups and zeroing out the others. GS-LoRA is effective, parameter-efficient, data-efficient, and easy to implement. We conduct extensive experiments on face recognition, object detection and image classification and demonstrate that GS-LoRA manages to forget specific classes with minimal impact on other classes. Codes will be released on \url{https://github.com/bjzhb666/GS-LoRA}., Comment: Accepted by CVPR 2024, latest version ahead of carema ready version

Low-rank adaptation (LoRA) is one of the most popular task-specific parameter-efficient fine-tuning (PEFT) methods on pre-trained language models for its good performance and computational efficiency. LoRA injects a product of two trainable rank decomposition matrices over the top of each frozen pre-trained model module. However, when applied in the setting of privacy-preserving federated learning (FL), LoRA may become unstable due to the following facts: 1) the effects of data heterogeneity and multi-step local updates are non-negligible, 2) additive noise enforced on updating gradients to guarantee differential privacy (DP) can be amplified and 3) the final performance is susceptible to hyper-parameters. A key factor leading to these phenomena is the discordance between jointly optimizing the two low-rank matrices by local clients and separately aggregating them by the central server. Thus, this paper proposes an efficient and effective version of LoRA, Federated Freeze A LoRA (FFA-LoRA), to alleviate these challenges and further halve the communication cost of federated fine-tuning LLMs. The core idea of FFA-LoRA is to fix the randomly initialized non-zero matrices and only fine-tune the zero-initialized matrices. Compared to LoRA, FFA-LoRA is motivated by practical and theoretical benefits in privacy-preserved FL. Our experiments demonstrate that FFA-LoRA provides more consistent performance with better computational efficiency over vanilla LoRA in various FL tasks., Comment: published at ICLR 2024, full paper 17 pages

We study a specific class of CFTs that involve coupled free fermions, arising from parafermion CFTs and lattice constructions. We analyse their representation spaces and the underlying exclusion statistics of coupled free fermions using specific bases. In one particular case, we reveal an unexpected connection between the coset construction of parafermions, the lattice construction, and and orbifold thereof. This connection is supported by proving a range of character identities within the context of coupled free fermions. Simultaneously, we obtain explicit expressions of certain string functions in terms of Dedekind eta functions.

In this paper we will prove a series of $q$-identities suggested by the realisation of certain conformal field theories by so-called `coupled free fermions'. We will consider $q$-series arising from coupled free fermions constructed by the parafermion coset construction as well as from scaled root lattices, and some interesting relations between the two.

Magnetic cooling based on the magnetocaloric effect is a promising solid-state refrigeration technology for a wide range of applications in different temperature ranges. Previous studies have mostly focused on near room temperature (300 K) and cryogenic temperature (< 10 K) ranges, while important applications such as hydrogen liquefaction call for efficient magnetic refrigerants for the intermediate temperature 10K to 100 K. For efficient use in this range, new magnetocaloric materials with matching Curie temperatures need to be discovered, while conventional experimental approaches are typically time-consuming and expensive. Here, we report a computational material discovery pipeline based on a materials database containing more than 6000 entries auto-generated by extracting reported material properties from literature using a large language model. We then use this database to train a machine learning model that can efficiently predict magnetocaloric properties of materials based on their chemical composition. We further verify the magnetocaloric properties of predicted compounds using ab initio atomistic spin dynamics simulations to close the loop for computational material discovery. Using this approach, we identify 11 new promising magnetocaloric materials for the target temperature range. Our work demonstrates the potential of combining large language models, machine learning, and ab initio simulations to efficiently discover new functional materials.

Understanding social interactions involving both verbal and non-verbal cues is essential for effectively interpreting social situations. However, most prior works on multimodal social cues focus predominantly on single-person behaviors or rely on holistic visual representations that are not aligned to utterances in multi-party environments. Consequently, they are limited in modeling the intricate dynamics of multi-party interactions. In this paper, we introduce three new challenging tasks to model the fine-grained dynamics between multiple people: speaking target identification, pronoun coreference resolution, and mentioned player prediction. We contribute extensive data annotations to curate these new challenges in social deduction game settings. Furthermore, we propose a novel multimodal baseline that leverages densely aligned language-visual representations by synchronizing visual features with their corresponding utterances. This facilitates concurrently capturing verbal and non-verbal cues pertinent to social reasoning. Experiments demonstrate the effectiveness of the proposed approach with densely aligned multimodal representations in modeling fine-grained social interactions. Project website: https://sangmin-git.github.io/projects/MMSI., Comment: CVPR 2024 Oral