Search

The integration of uncertainty sets has diverse applications but also presents challenges, particularly when only initial sets and their control levels are available, along with potential dependencies. Examples include merging confidence sets from different distributed sites with communication constraints, as well as combining conformal prediction sets generated by different learning algorithms or data splits. In this article, we introduce an efficient and flexible Synthetic, Aggregation, and Test inversion (SAT) approach to merge various potentially dependent uncertainty sets into a single set. The proposed method constructs a novel class of synthetic test statistics, aggregates them, and then derives merged sets through test inversion. Our approach leverages the duality between set estimation and hypothesis testing, ensuring reliable coverage in dependent scenarios. The procedure is data-light, meaning it relies solely on initial sets and control levels without requiring raw data, and it adapts to any user-specified initial uncertainty sets, accommodating potentially varying coverage levels. Theoretical analyses and numerical experiments confirm that SAT provides finite-sample coverage guarantees and achieves small set sizes.

In this paper, we present a computational algorithm for constructing all boundaries and defects of topological generalized Pauli stabilizer codes in two spatial dimensions. Utilizing the operator algebra formalism, we establish a one-to-one correspondence between the topological data-such as anyon types, fusion rules, topological spins, and braiding statistics-of (2+1)D bulk stabilizer codes and (1+1)D boundary anomalous subsystem codes. To make the operator algebra computationally accessible, we adapt Laurent polynomials and convert the tasks into matrix operations, e.g., the Hermite normal form for obtaining boundary anyons and the Smith normal form for determining fusion rules. This approach enables computers to automatically generate all possible gapped boundaries and defects for topological Pauli stabilizer codes through boundary anyon condensation and topological order completion. This streamlines the analysis of surface codes and associated logical operations for fault-tolerant quantum computation. Our algorithm applies to $Z_d$ qudits, including both prime and nonprime $d$, thus enabling the exploration of topological quantum codes beyond toric codes. We have applied the algorithm and explicitly demonstrated the lattice constructions of 2 boundaries and 6 defects in the $Z_2$ toric code, 3 boundaries and 22 defects in the $Z_4$ toric code, 1 boundary and 2 defects in the double semion code, 1 boundary and 22 defects in the six-semion code, 6 boundaries and 270 defects in the color code, and 6 defects in the anomalous three-fermion code. In addition, we investigate the boundaries of two specific bivariate bicycle codes within a family of low-density parity-check (LDPC) codes. We demonstrate that their topological orders are equivalent to 8 and 10 copies of $Z_2$ toric codes, with anyons restricted to move by 12 and 1023 lattice sites in the square lattice, respectively., Comment: 36+17 pages, 41 figures; v2: typos fixed, figures updated; v3: format changed

As one of the most popular and sought-after generative models in the recent years, diffusion models have sparked the interests of many researchers and steadily shown excellent advantage in various generative tasks such as image synthesis, video generation, molecule design, 3D scene rendering and multimodal generation, relying on their dense theoretical principles and reliable application practices. The remarkable success of these recent efforts on diffusion models comes largely from progressive design principles and efficient architecture, training, inference, and deployment methodologies. However, there has not been a comprehensive and in-depth review to summarize these principles and practices to help the rapid understanding and application of diffusion models. In this survey, we provide a new efficiency-oriented perspective on these existing efforts, which mainly focuses on the profound principles and efficient practices in architecture designs, model training, fast inference and reliable deployment, to guide further theoretical research, algorithm migration and model application for new scenarios in a reader-friendly way. \url{https://github.com/ponyzym/Efficient-DMs-Survey}

Diffusion models have achieved excellent success in solving inverse problems due to their ability to learn strong image priors, but existing approaches require a large training dataset of images that should come from the same distribution as the test dataset. When the training and test distributions are mismatched, artifacts and hallucinations can occur in reconstructed images due to the incorrect priors. In this work, we systematically study out of distribution (OOD) problems where a known training distribution is first provided. We first study the setting where only a single measurement obtained from the unknown test distribution is available. Next we study the setting where a very small sample of data belonging to the test distribution is available, and our goal is still to reconstruct an image from a measurement that came from the test distribution. In both settings, we use a patch-based diffusion prior that learns the image distribution solely from patches. Furthermore, in the first setting, we include a self-supervised loss that helps the network output maintain consistency with the measurement. Extensive experiments show that in both settings, the patch-based method can obtain high quality image reconstructions that can outperform whole-image models and can compete with methods that have access to large in-distribution training datasets. Furthermore, we show how whole-image models are prone to memorization and overfitting, leading to artifacts in the reconstructions, while a patch-based model can resolve these issues.

Recent approaches for visually-rich document understanding (VrDU) uses manually annotated semantic groups, where a semantic group encompasses all semantically relevant but not obviously grouped words. As OCR tools are unable to automatically identify such grouping, we argue that current VrDU approaches are unrealistic. We thus introduce a new variant of the VrDU task, real-world visually-rich document understanding (ReVrDU), that does not allow for using manually annotated semantic groups. We also propose a new method, ReLayout, compliant with the ReVrDU scenario, which learns to capture semantic grouping through arranging words and bringing the representations of words that belong to the potential same semantic group closer together. Our experimental results demonstrate the performance of existing methods is deteriorated with the ReVrDU task, while ReLayout shows superiour performance.

Current LLM customization typically relies on two deployment strategies: closed-source APIs, which require users to upload private data to external servers, and open-weight models, which allow local fine-tuning but pose misuse risks. In this position paper, we argue that (1) deploying closed-source LLMs within user-controlled infrastructure (\textit{on-premises deployment}) enhances data privacy and mitigates misuse risks, and (2) a well-designed on-premises deployment must ensure model confidentiality -- by preventing model theft -- and offer privacy-preserving customization. Prior research on small models has explored securing only the output layer within hardware-secured devices to balance confidentiality and customization efficiency. However, we show that this approach is insufficient for defending large-scale LLMs against distillation attacks. We therefore introduce a {semi-open deployment framework} that secures only a few, carefully chosen layers, achieving distillation resistance comparable to fully secured models while preserving fine-tuning flexibility. Through extensive experiments, we show that securing bottom layers significantly reduces functional extraction risks. Our findings demonstrate that privacy and confidentiality can coexist, paving the way for secure on-premises AI deployment that balances usability and protection., Comment: 8 pages for main content of the paper

Transferable targeted adversarial attacks (TTAs) against deep neural networks have been proven significantly more challenging than untargeted ones, yet they remain relatively underexplored. This paper sheds new light on performing highly efficient yet transferable targeted attacks leveraging the simple gradient-based baseline. Our research underscores the critical importance of image transformations within gradient calculations, marking a shift from the prevalent emphasis on loss functions to address the gradient vanishing problem. Moreover, we have developed two effective blind estimators that facilitate the design of transformation strategies to enhance targeted transferability under black-box conditions. The adversarial examples' self-transferability to geometric transformations has been identified as strongly correlated with their black-box transferability, featuring these basic operations as potent yet overlapped proxies for facilitating targeted transferability. The surrogate self-alignment assessments further highlight simple scaling transformation's exceptional efficacy, which rivals that of most advanced methods. Building on these insights, we introduce a scaling-centered transformation strategy termed Strong, Self-transferable, faSt, and Simple Scale Transformation (S4ST) to enhance transferable targeted attacks. In experiments conducted on the ImageNet-Compatible benchmark dataset, our proposed S4ST attains a SOTA average targeted transfer success rate across various challenging black-box models, outperforming the previous leading method by over 14% while requiring only 25% of the execution time. Additionally, our approach eclipses SOTA attacks considerably and exhibits remarkable effectiveness against real-world APIs. This work marks a significant leap forward in TTAs, revealing the realistic threats they pose and providing a practical generation method for future research., Comment: 16 pages, 18 figures

The right to be forgotten mandates that machine learning models enable the erasure of a data owner's data and information from a trained model. Removing data from the dataset alone is inadequate, as machine learning models can memorize information from the training data, increasing the potential privacy risk to users. To address this, multiple machine unlearning techniques have been developed and deployed. Among them, approximate unlearning is a popular solution, but recent studies report that its unlearning effectiveness is not fully guaranteed. Another approach, exact unlearning, tackles this issue by discarding the data and retraining the model from scratch, but at the cost of considerable computational and memory resources. However, not all devices have the capability to perform such retraining. In numerous machine learning applications, such as edge devices, Internet-of-Things (IoT), mobile devices, and satellites, resources are constrained, posing challenges for deploying existing exact unlearning methods. In this study, we propose a Constraint-aware Adaptive Exact Unlearning System at the network Edge (CAUSE), an approach to enabling exact unlearning on resource-constrained devices. Aiming to minimize the retrain overhead by storing sub-models on the resource-constrained device, CAUSE innovatively applies a Fibonacci-based replacement strategy and updates the number of shards adaptively in the user-based data partition process. To further improve the effectiveness of memory usage, CAUSE leverages the advantage of model pruning to save memory via compression with minimal accuracy sacrifice. The experimental results demonstrate that CAUSE significantly outperforms other representative systems in realizing exact unlearning on the resource-constrained device by 9.23%-80.86%, 66.21%-83.46%, and 5.26%-194.13% in terms of unlearning speed, energy consumption, and accuracy., Comment: Accepted to IEEE Symposium on Security and Privacy 2025 (Oakland 2025)

Recently, plenty research has been done on discovering the role of energy dissipation in biological networks, most of which focus on the relationship of dissipation and functionality. However, the development of networks science urged us to fathom the systematic architecture of biological networks and their evolutionary advantages. We found the dissipation of biological dissipative networks is highly related to their structure. By interrogating these well-adapted networks, we find that the energy producing module is relatively isolated in all situations. We applied evolutionary simulation and analysis on premature networks of classic dissipative networks, namely kinetic proofreading, activator-inhibitor oscillator and two typical adaptative response models. We found despite that selection was imposed merely on the network function, the networks tended to decouple high energy molecules as fuels from the functional module, to achieve higher overall dissipation during the course of evolution. Furthermore, we find that decoupled fuel modules can increase the robustness of the networks towards parameter or structure perturbations. We provide theoretical analysis on the kinetic proofreading networks and the general case of energy-driven networks. We find fuel decoupling can guarantee higher dissipation and, in most cases when considering dissipative networks, higher performance. We conclude that fuel decoupling is an evolutionary outcome and bears benefits during evolution., Comment: 12 pages, 6 figures

The rapid advancement of scientific progress requires innovative tools that can accelerate knowledge discovery. Although recent AI methods, particularly large language models (LLMs), have shown promise in tasks such as hypothesis generation and experimental design, they fall short of replicating the collaborative nature of real-world scientific practices, where diverse experts work together in teams to tackle complex problems. To address the limitations, we propose an LLM-based multi-agent system, i.e., Virtual Scientists (VirSci), designed to mimic the teamwork inherent in scientific research. VirSci organizes a team of agents to collaboratively generate, evaluate, and refine research ideas. Through comprehensive experiments, we demonstrate that this multi-agent approach outperforms the state-of-the-art method in producing novel scientific ideas. We further investigate the collaboration mechanisms that contribute to its tendency to produce ideas with higher novelty, offering valuable insights to guide future research and illuminating pathways toward building a robust system for autonomous scientific discovery. The code is available at https://github.com/open-sciencelab/Virtual-Scientists.

We consider the finite generation property for cohomology algebra of pointed finite tensor categories via de-equivariantization and exact sequence of finite tensor categories. As a result, we prove that all coradically graded pointed finite tensor categories over abelian groups have finitely generated cohomology., Comment: 11 pages, comments welcome

Supervised fine-tuning (SFT) is crucial for aligning Large Language Models (LLMs) with human instructions. The primary goal during SFT is to select a small yet representative subset of training data from the larger pool, such that fine-tuning with this subset achieves results comparable to or even exceeding those obtained using the entire dataset. However, most existing data selection techniques are designed for small-scale data pools, which fail to meet the demands of real-world SFT scenarios. In this paper, we replicated several self-scoring methods those that do not rely on external model assistance on two million scale datasets, and found that nearly all methods struggled to significantly outperform random selection when dealing with such large-scale data pools. Moreover, our comparisons suggest that, during SFT, diversity in data selection is more critical than simply focusing on high quality data. We also analyzed the limitations of several current approaches, explaining why they perform poorly on large-scale datasets and why they are unsuitable for such contexts. Finally, we found that filtering data by token length offers a stable and efficient method for improving results. This approach, particularly when training on long text data, proves highly beneficial for relatively weaker base models, such as Llama3.

Length extrapolation algorithms based on Rotary position embedding (RoPE) have shown promising results in extending the context length of language models. However, understanding how position embedding can capture longer-range contextual information remains elusive. Based on the intuition that different dimensions correspond to different frequency of changes in RoPE encoding, we conducted a dimension-level analysis to investigate the correlation between a hidden dimension of an attention head and its contribution to capturing long-distance dependencies. Using our correlation metric, we identified a particular type of attention heads, which we named Positional Heads, from various length-extrapolated models. These heads exhibit a strong focus on long-range information interaction and play a pivotal role in long input processing, as evidence by our ablation. We further demonstrate the correlation between the efficiency of length extrapolation and the extension of the high-dimensional attention allocation of these heads. The identification of Positional Heads provides insights for future research in long-text comprehension., Comment: Accepted to EMNLP 2024 Findings

The salient multimodal capabilities and interactive experience of GPT-4o highlight its critical role in practical applications, yet it lacks a high-performing open-source counterpart. In this paper, we introduce Baichuan-omni, the first open-source 7B Multimodal Large Language Model (MLLM) adept at concurrently processing and analyzing modalities of image, video, audio, and text, while delivering an advanced multimodal interactive experience and strong performance. We propose an effective multimodal training schema starting with 7B model and proceeding through two stages of multimodal alignment and multitask fine-tuning across audio, image, video, and text modal. This approach equips the language model with the ability to handle visual and audio data effectively. Demonstrating strong performance across various omni-modal and multimodal benchmarks, we aim for this contribution to serve as a competitive baseline for the open-source community in advancing multimodal understanding and real-time interaction.

Ruling a wide range of phenomena, dipole-dipole interactions (DDI) are typically constrained to the short range due to their rapid decay with the increasing dipole separations, limiting the performance in long-range applications. By judiciously designing the photonic structures that control the two-point Green's functions of the electromagnetic environment, the spontaneous emission of quantum emitters (luminescence) and their interactions (e.g., F\"orster energy transfer) can be conveniently tuned. In this paper, we designed a matched nanoparticle-on-mirror antenna pair with enhanced DDI guided by surface plasmon polaritons confined to the metal substrate, which ensures concentrated and enhanced interaction over long ranges of tens of wavelengths. The long-range ($\sim 10 \lambda$) DDI between donor-acceptor emitters is enhanced by $6\times 10^{3}$ times respective to bare gold film, and $4.4\times 10^{4}$ times respective to vacuum. Our result provides a promising testbed for investigating long-range DDI phenomena on the nanoscale.

While advancements in NLP have significantly improved the performance of Large Language Models (LLMs) on tasks requiring vertical thinking, their lateral thinking capabilities remain under-explored and challenging to measure due to the complexity of assessing creative thought processes and the scarcity of relevant data. To address these challenges, we introduce SPLAT, a benchmark leveraging Situation Puzzles to evaluate and elicit LAteral Thinking of LLMs. This benchmark, containing 975 graded situation puzzles across three difficulty levels, employs a new multi-turn player-judge framework instead of the traditional model-based evaluation, which often necessitates a stronger evaluation model. This framework simulates an interactive game where the model (player) asks the evaluation model (judge) questions about an incomplete story to infer the full scenario. The judge answers based on a detailed reference scenario or evaluates if the player's predictions align with the reference one. This approach lessens dependence on more robust evaluation models, enabling the assessment of state-of-the-art LLMs. The experiments demonstrate that a robust evaluation model, such as WizardLM-2, closely matches human judgements in both intermediate question-answering and final scenario accuracy, achieving over 80% agreement-similar to the agreement levels among humans. Furthermore, applying data and reasoning processes from our benchmark to other lateral thinking-related benchmarks, e.g., RiddleSense and BrainTeaser, leads to performance enhancements. This suggests that our benchmark effectively evaluates and elicits the lateral thinking abilities of LLMs. Code is available at: https://github.com/chenqi008/LateralThinking., Comment: Accepted by NeurIPS 2024

In various scientific and engineering fields, the primary research areas have revolved around physics-based dynamical systems modeling and data-driven time series analysis. According to the embedding theory, dynamical systems and time series can be mutually transformed using observation functions and physical reconstruction techniques. Based on this, we propose Embedding Duality Theory, where the parameterized embedding layer essentially provides a linear estimation of the non-linear time series dynamics. This theory enables us to bypass the parameterized embedding layer and directly employ physical reconstruction techniques to acquire a data embedding representation. Utilizing physical priors results in a 10X reduction in parameters, a 3X increase in speed, and maximum performance boosts of 18% in expert, 22% in few-shot, and 53\% in zero-shot tasks without any hyper-parameter tuning. All methods are encapsulated as a plug-and-play module

This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel "Memory Pointer Prompting" mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding., Comment: Technical Report

In this paper, we approach an overlooked yet critical task Graph2Image: generating images from multimodal attributed graphs (MMAGs). This task poses significant challenges due to the explosion in graph size, dependencies among graph entities, and the need for controllability in graph conditions. To address these challenges, we propose a graph context-conditioned diffusion model called InstructG2I. InstructG2I first exploits the graph structure and multimodal information to conduct informative neighbor sampling by combining personalized page rank and re-ranking based on vision-language features. Then, a Graph-QFormer encoder adaptively encodes the graph nodes into an auxiliary set of graph prompts to guide the denoising process of diffusion. Finally, we propose graph classifier-free guidance, enabling controllable generation by varying the strength of graph guidance and multiple connected edges to a node. Extensive experiments conducted on three datasets from different domains demonstrate the effectiveness and controllability of our approach. The code is available at https://github.com/PeterGriffinJin/InstructG2I., Comment: 16 pages

Neural operators, which learn mappings between the function spaces, have been applied to solve boundary value problems in various ways, including learning mappings from the space of the forcing terms to the space of the solutions with the substantial requirements of data pairs. In this work, we present a data-free neural operator integrated with physics, which learns the Green kernel directly. Our method proceeds in three steps: 1. The governing equations for the Green's function are reformulated into an interface problem, where the delta Dirac function is removed; 2. The interface problem is embedded in a lifted space of higher-dimension to handle the jump in the derivative, but still solved on a two-dimensional surface without additional sampling cost; 3. Deep neural networks are employed to address the curse of dimensionality caused by this lifting operation. The approximate Green's function obtained through our approach is then used to construct preconditioners for the linear systems allowed by its mathematical properties. Furthermore, the spectral bias of it revealed through both theoretical analysis and numerical validation contrasts with the smoothing effects of traditional iterative solvers, which motivates us to propose a hybrid iterative method that combines these two solvers. Numerical experiments demonstrate the effectiveness of our approximate Green's function in accelerating iterative methods, proving fast convergence for solving indefinite problems even involving discontinuous coefficients.

The soft-argmax operation is widely adopted in neural network-based stereo matching methods to enable differentiable regression of disparity. However, network trained with soft-argmax is prone to being multimodal due to absence of explicit constraint to the shape of the probability distribution. Previous methods leverages Laplacian distribution and cross-entropy for training but failed to effectively improve the accuracy and even compromises the efficiency of the network. In this paper, we conduct a detailed analysis of the previous distribution-based methods and propose a novel supervision method for stereo matching, Sampling-Gaussian. We sample from the Gaussian distribution for supervision. Moreover, we interpret the training as minimizing the distance in vector space and propose a combined loss of L1 loss and cosine similarity loss. Additionally, we leveraged bilinear interpolation to upsample the cost volume. Our method can be directly applied to any soft-argmax-based stereo matching method without a reduction in efficiency. We have conducted comprehensive experiments to demonstrate the superior performance of our Sampling-Gaussian. The experimental results prove that we have achieved better accuracy on five baseline methods and two datasets. Our method is easy to implement, and the code is available online., Comment: TL;DR: A novel Gaussian distribution-based supervision method for stereo matching. Implemented with five baseline methods and achieves notable improvement. Main content, 10 pages. conference submission

Homomorphic secret sharing (HSS) enables two servers to locally perform functions on encrypted data directly and obtain the results in the form of shares. A Paillier-based HSS solution seamlessly achieves multiplicative homomorphism and consumes less communication costs. Unfortunately, existing Paillier-based HSS schemes suffer from a large private key size, potential calculation error, expensive computation and storage overhead, and only valid on linear operations (e.g., addition and multiplication). To this end, inspired by the Paillier cryptosystem with fast encryption and decryption, we propose MORSE, an efficient homomorphic secret sharing scheme enabling non-linear operation, which enjoys a small key size, no calculation error and low overhead. In terms of functions, MORSE supports addition, subtraction, multiplication, scalar-multiplication, and comparison. Particularly, we carefully design two conversion protocols achieving the mutual conversion between one Paillier ciphertext and two secret shares, which allows MORSE to continuously perform the above operations. Rigorous analyses demonstrate that MORSE securely outputs correct results. Experimental results show that MORSE makes a runtime improvement of up to 9.3 times in terms of secure multiplication, and a communication costs reduction of up to 16.6% in secure comparison, compared to the state-of-the-art.

In the rapidly evolving field of deep learning, specialized models have driven significant advancements in tasks such as computer vision and natural language processing. However, this specialization leads to a fragmented ecosystem where models lack the adaptability for broader applications. To overcome this, we introduce AutoFusion, an innovative framework that fuses distinct model parameters(with the same architecture) for multi-task learning without pre-trained checkpoints. Using an unsupervised, end-to-end approach, AutoFusion dynamically permutes model parameters at each layer, optimizing the combination through a loss-minimization process that does not require labeled data. We validate AutoFusion's effectiveness through experiments on commonly used benchmark datasets, demonstrating superior performance over established methods like Weight Interpolation, Git Re-Basin, and ZipIt. Our framework offers a scalable and flexible solution for model integration, positioning it as a powerful tool for future research and practical applications., Comment: Under review

Skeleton-based Temporal Action Segmentation involves the dense action classification of variable-length skeleton sequences. Current approaches primarily apply graph-based networks to extract framewise, whole-body-level motion representations, and use one-hot encoded labels for model optimization. However, whole-body motion representations do not capture fine-grained part-level motion representations and the one-hot encoded labels neglect the intrinsic semantic relationships within the language-based action definitions. To address these limitations, we propose a novel method named Language-assisted Human Part Motion Representation Learning (LPL), which contains a Disentangled Part Motion Encoder (DPE) to extract dual-level (i.e., part and whole-body) motion representations and a Language-assisted Distribution Alignment (LDA) strategy for optimizing spatial relations within representations. Specifically, after part-aware skeleton encoding via DPE, LDA generates dual-level action descriptions to construct a textual embedding space with the help of a large-scale language model. Then, LDA motivates the alignment of the embedding space between text descriptions and motions. This alignment allows LDA not only to enhance intra-class compactness but also to transfer the language-encoded semantic correlations among actions to skeleton-based motion learning. Moreover, we propose a simple yet efficient Semantic Offset Adapter to smooth the cross-domain misalignment. Our experiments indicate that LPL achieves state-of-the-art performance across various datasets (e.g., +4.4\% Accuracy, +5.6\% F1 on the PKU-MMD dataset). Moreover, LDA is compatible with existing methods and improves their performance (e.g., +4.8\% Accuracy, +4.3\% F1 on the LARa dataset) without additional inference costs., Comment: This work has been submitted to the IEEE for possible publication

The solvability of equilibrium Riccati equations (EREs) plays a central role in the study of time-inconsistent stochastic linear-quadratic optimal control problems, because it paves the way to constructing a closed-loop equilibrium strategy. Under the standard conditions, Yong [29] established its well-posedness by introducing the well-known multi-person differential game method. However, this method depends on the dynamic programming principle (DPP) of the sophisticated problems on every subinterval, and thus is essentially a control theory approach. In this paper, we shall give a new and more direct proof, in which the DPP is no longer needed. We first establish a priori estimates for the ERE in the case of smooth coefficients. Using this estimate, we then demonstrate both the local and global solvability of the ERE by constructing an appropriate Picard iteration sequence, which actually provides a numerical algorithm. Additionally, a mollification method is employed to handle the case with non-smooth coefficients.

Information comes in diverse modalities. Multimodal native AI models are essential to integrate real-world information and deliver comprehensive understanding. While proprietary multimodal native models exist, their lack of openness imposes obstacles for adoptions, let alone adaptations. To fill this gap, we introduce Aria, an open multimodal native model with best-in-class performance across a wide range of multimodal, language, and coding tasks. Aria is a mixture-of-expert model with 3.9B and 3.5B activated parameters per visual token and text token, respectively. It outperforms Pixtral-12B and Llama3.2-11B, and is competitive against the best proprietary models on various multimodal tasks. We pre-train Aria from scratch following a 4-stage pipeline, which progressively equips the model with strong capabilities in language understanding, multimodal understanding, long context window, and instruction following. We open-source the model weights along with a codebase that facilitates easy adoptions and adaptations of Aria in real-world applications.

Retrieval-augmented generation (RAG) empowers large language models (LLMs) to utilize external knowledge sources. The increasing capacity of LLMs to process longer input sequences opens up avenues for providing more retrieved information, to potentially enhance the quality of generated outputs. It is plausible to assume that a larger retrieval set would contain more relevant information (higher recall), that might result in improved performance. However, our empirical findings demonstrate that for many long-context LLMs, the quality of generated output initially improves first, but then subsequently declines as the number of retrieved passages increases. This paper investigates this phenomenon, identifying the detrimental impact of retrieved "hard negatives" as a key contributor. To mitigate this and enhance the robustness of long-context LLM-based RAG, we propose both training-free and training-based approaches. We first showcase the effectiveness of retrieval reordering as a simple yet powerful training-free optimization. Furthermore, we explore training-based methods, specifically RAG-specific implicit LLM fine-tuning and RAG-oriented fine-tuning with intermediate reasoning, demonstrating their capacity for substantial performance gains. Finally, we conduct a systematic analysis of design choices for these training-based methods, including data distribution, retriever selection, and training context length., Comment: 34 pages

As Large Language Models (LLMs) excel across tasks and specialized domains, scaling LLMs based on existing models has garnered significant attention, which faces the challenge of decreasing performance when combining disparate models. Various techniques have been proposed for the aggregation of pre-trained LLMs, including model merging, Mixture-of-Experts, and stacking. Despite their merits, a comprehensive comparison and synergistic application of them to a diverse model zoo is yet to be adequately addressed. In light of this research gap, this paper introduces Model-GLUE, a holistic LLM scaling guideline. First, our work starts with a benchmarking of existing LLM scaling techniques, especially selective merging, and variants of mixture. Utilizing the insights from the benchmark results, we formulate an optimal strategy for the selection and aggregation of a heterogeneous model zoo characterizing different architectures and initialization.Our methodology involves the clustering of mergeable models and optimal merging strategy selection, and the integration of clusters through a model mixture. Finally, evidenced by our experiments on a diverse Llama-2-based model zoo, Model-GLUE shows an average performance enhancement of 5.61%, achieved without additional training. Codes are available at: https://github.com/Model-GLUE/Model-GLUE., Comment: 24 pages, 4 figures, accepted to NeurIPS 2024 Datasets and Benchmarks Track

Given the increasing adoption of modern AI-enabled control systems, ensuring their safety and reliability has become a critical task in software testing. One prevalent approach to testing control systems is falsification, which aims to find an input signal that causes the control system to violate a formal safety specification using optimization algorithms. However, applying falsification to AI-enabled control systems poses two significant challenges: (1)~it requires the system to execute numerous candidate test inputs, which can be time-consuming, particularly for systems with AI models that have many parameters, and (2)~multiple safety requirements are typically defined as a conjunctive specification, which is difficult for existing falsification approaches to comprehensively cover. This paper introduces Synthify, a falsification framework tailored for AI-enabled control systems. Our approach performs falsification in a two-phase process. At the start, Synthify synthesizes a program that implements one or a few linear controllers to serve as a proxy for the AI controller. This proxy program mimics the AI controller's functionality but is computationally more efficient. Then, Synthify employs the $\epsilon$-greedy strategy to sample a promising sub-specification from the conjunctive safety specification. It then uses a Simulated Annealing-based falsification algorithm to find violations of the sampled sub-specification for the control system. To evaluate Synthify, we compare it to PSY-TaLiRo, a state-of-the-art and industrial-strength falsification tool, on 8 publicly available control systems. On average, Synthify achieves a 83.5% higher success rate in falsification compared to PSY-TaLiRo with the same budget of falsification trials. The safety violations found by Synthify are also more diverse than those found by PSY-TaLiRo, covering 137.7% more sub-specifications., Comment: Accepted by ACM Transactions on Software Engineering and Methodology (TOSEM), 35 pages

The four-dimensional variational data assimilation (4D-Var) has emerged as an important methodology, widely used in numerical weather prediction, oceanographic modeling, and climate forecasting. Classical unconstrained gradient-based algorithms often struggle with local minima, making their numerical performance highly sensitive to the initial guess. In this study, we exploit the separable structure of the 4D-Var problem to propose a practical variant of the alternating direction method of multipliers (ADMM), referred to as the linearized multi-block ADMM with regularization. Unlike classical first-order optimization methods that primarily focus on initial conditions, our approach derives the Euler-Lagrange equation for the entire dynamical system, enabling more comprehensive and effective utilization of observational data. When the initial condition is poorly chosen, the arg min operation steers the iteration towards the observational data, thereby reducing sensitivity to the initial guess. The quadratic subproblems further simplify the solution process, while the parallel structure enhances computational efficiency, especially when utilizing modern hardware. To validate our approach, we demonstrate its superior performance using the Lorenz system, even in the presence of noisy observational data. Furthermore, we showcase the effectiveness of the linearized multi-block ADMM with regularization in solving the 4D-Var problems for the viscous Burgers' equation, across various numerical schemes, including finite difference, finite element, and spectral methods. Finally, we illustrate the recovery of dynamics under noisy observational data in a 2D turbulence scenario, particularly focusing on vorticity concentration, highlighting the robustness of our algorithm in handling complex physical phenomena., Comment: 26 pages, 16 figures, 1 table

Tokenizer is an essential component for large language models (LLMs), and a tokenizer with a high compression rate can improve the model's representation and processing efficiency. However, the tokenizer cannot ensure high compression rate in all scenarios, and an increase in the average input and output lengths will increases the training and inference costs of the model. Therefore, it is crucial to find ways to improve the model's efficiency with minimal cost while maintaining the model's performance. In this work, we propose a method to improve model representation and processing efficiency by replacing the tokenizers of LLMs. We propose replacing and reinitializing the parameters of the model's input and output layers with the parameters of the original model, and training these parameters while keeping other parameters fixed. We conducted experiments on different LLMs, and the results show that our method can maintain the performance of the model after replacing the tokenizer, while significantly improving the decoding speed for long texts.

The context window of large language models (LLMs) has been extended significantly in recent years. However, while the context length that the LLM can process has grown, the capability of the model to accurately reason over that context degrades noticeably. This occurs because modern LLMs often become overwhelmed by the vast amount of information in the context; when answering questions, the model must identify and reason over relevant evidence sparsely distributed throughout the text. To alleviate the challenge of long-context reasoning, we develop a retrieve-then-reason framework, enabling LLMs to reason over relevant evidence collected during an intermediate retrieval step. We find that modern LLMs struggle to accurately retrieve relevant facts and instead, often hallucinate "retrieved facts", resulting in flawed reasoning and the production of incorrect answers. To address these issues, we introduce ALR$^2$, a method that augments the long-context reasoning capability of LLMs via an explicit two-stage procedure, i.e., aligning LLMs with the objectives of both retrieval and reasoning. We demonstrate the efficacy of ALR$^2$ for mitigating performance degradation in long-context reasoning tasks. Through extensive experiments on long-context QA benchmarks, we find our method to outperform competitive baselines by large margins, achieving at least 8.4 and 7.9 EM gains on the long-context versions of HotpotQA and SQuAD datasets, respectively.

Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks., Comment: Preprint

Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models., Comment: CV/ML

Diffusion transformers have been widely adopted for text-to-image synthesis. While scaling these models up to billions of parameters shows promise, the effectiveness of scaling beyond current sizes remains underexplored and challenging. By explicitly exploiting the computational heterogeneity of image generations, we develop a new family of Mixture-of-Experts (MoE) models (EC-DIT) for diffusion transformers with expert-choice routing. EC-DIT learns to adaptively optimize the compute allocated to understand the input texts and generate the respective image patches, enabling heterogeneous computation aligned with varying text-image complexities. This heterogeneity provides an efficient way of scaling EC-DIT up to 97 billion parameters and achieving significant improvements in training convergence, text-to-image alignment, and overall generation quality over dense models and conventional MoE models. Through extensive ablations, we show that EC-DIT demonstrates superior scalability and adaptive compute allocation by recognizing varying textual importance through end-to-end training. Notably, in text-to-image alignment evaluation, our largest models achieve a state-of-the-art GenEval score of 71.68% and still maintain competitive inference speed with intuitive interpretability.

Long-Form Question Answering (LFQA) refers to generating in-depth, paragraph-level responses to open-ended questions. Although lots of LFQA methods are developed, evaluating LFQA effectively and efficiently remains challenging due to its high complexity and cost. Therefore, there is no standard benchmark for LFQA evaluation till now. To address this gap, we make the first attempt by proposing a well-constructed, reference-based benchmark named Chinese exAmination for LFQA Evaluation (CALF), aiming to rigorously assess the performance of automatic evaluation metrics for LFQA. The CALF benchmark is derived from Chinese examination questions that have been translated into English. It includes up to 1476 examples consisting of knowledge-intensive and nuanced responses. Our evaluation comprises three different settings to ana lyze the behavior of automatic metrics comprehensively. We conducted extensive experiments on 7 traditional evaluation metrics, 3 prompt-based metrics, and 3 trained evaluation metrics, and tested on agent systems for the LFQA evaluation. The results reveal that none of the current automatic evaluation metrics shows comparable performances with humans, indicating that they cannot capture dense information contained in long-form responses well. In addition, we provide a detailed analysis of the reasons why automatic evaluation metrics fail when evaluating LFQA, offering valuable insights to advance LFQA evaluation systems. Dataset and associated codes can be accessed at our GitHub repository.

We propose a polynomial-time algorithm for preparing the Gibbs state of the two-dimensional toric code Hamiltonian at any temperature, starting from any initial condition, significantly improving upon prior estimates that suggested exponential scaling with inverse temperature. Our approach combines the Lindblad dynamics using a local Davies generator with simple global jump operators to enable efficient transitions between logical sectors. We also prove that the Lindblad dynamics with a digitally implemented low temperature local Davies generator is able to efficiently drive the quantum state towards the ground state manifold. Despite this progress, we explain why protecting quantum information in the 2D toric code with passive dynamics remains challenging.

Understanding and predicting pedestrian crossing behavioral intention is crucial for autonomous vehicles driving safety. Nonetheless, challenges emerge when using promising images or environmental context masks to extract various factors for time-series network modeling, causing pre-processing errors or a loss in efficiency. Typically, pedestrian positions captured by onboard cameras are often distorted and do not accurately reflect their actual movements. To address these issues, GTransPDM -- a Graph-embedded Transformer with a Position Decoupling Module -- was developed for pedestrian crossing intention prediction by leveraging multi-modal features. First, a positional decoupling module was proposed to decompose the pedestrian lateral movement and simulate depth variations in the image view. Then, a graph-embedded Transformer was designed to capture the spatial-temporal dynamics of human pose skeletons, integrating essential factors such as position, skeleton, and ego-vehicle motion. Experimental results indicate that the proposed method achieves 92% accuracy on the PIE dataset and 87% accuracy on the JAAD dataset, with a processing speed of 0.05ms. It outperforms the state-of-the-art in comparison.

Harnessing quantum interference among bosons provides significant opportunities as bosons often carry longer coherence time than fermions. As an example of quantum interference, Fano resonance involving phonons or photons describes the coupling between discrete and continuous states, signified by an asymmetric spectral lineshape. Utilizing photon-based Fano resonance, molecule sensing with ultra-high sensitivity and ultrafast optical switching has been realized. However, phonon-based Fano resonance, which would expand the application space to a vaster regime, has been less exploited because of the weak coupling between discrete phonons with continuous states such as electronic continuum. In this work, we report the discovery of giant phonon-based Fano resonance in a graphene/2D Ag/SiC heterostructure. The Fano asymmetry, being proportional to the coupling strength, exceeds prior reports by two orders of magnitude. This Fano asymmetry arises from simultaneous frequency and lifetime matching between discrete and continuous phonons of SiC. The introduction of 2D Ag layers restructures SiC at the interface and facilitates resonant scattering to further enhance the Fano asymmetry, which is not achievable with conventional Ag thin films. With these unique properties, we demonstrated that the phonon-based Fano resonance can be used for ultrasensitive molecule detection at the single-molecule level. Our work highlights strong Fano resonance in the phononic system, opening avenues for engineering quantum interference based on bosons. Further, our findings provide opportunities for advancing phonon-related applications, including biochemical sensing, quantum transduction, and superconductor-based quantum computing.

We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development.

A primary challenge for visual-based Reinforcement Learning (RL) is to generalize effectively across unseen environments. Although previous studies have explored different auxiliary tasks to enhance generalization, few adopt image reconstruction due to concerns about exacerbating overfitting to task-irrelevant features during training. Perceiving the pre-eminence of image reconstruction in representation learning, we propose SMG (Separated Models for Generalization), a novel approach that exploits image reconstruction for generalization. SMG introduces two model branches to extract task-relevant and task-irrelevant representations separately from visual observations via cooperatively reconstruction. Built upon this architecture, we further emphasize the importance of task-relevant features for generalization. Specifically, SMG incorporates two additional consistency losses to guide the agent's focus toward task-relevant areas across different scenarios, thereby achieving free from overfitting. Extensive experiments in DMC demonstrate the SOTA performance of SMG in generalization, particularly excelling in video-background settings. Evaluations on robotic manipulation tasks further confirm the robustness of SMG in real-world applications.

Why do security cameras, sensors, and siri use cloud servers instead of on-board computation? The lack of very-low-power, high-performance chips greatly limits the ability to field untethered edge devices. We present the NV-1, a new low-power ASIC AI processor that greatly accelerates parallel processing (> 10X) with dramatic reduction in energy consumption (> 100X), via many parallel combined processor-memory units, i.e., a drastically non-von-Neumann architecture, allowing very large numbers of independent processing streams without bottlenecks due to typical monolithic memory. The current initial prototype fab arises from a successful co-development effort between algorithm- and software-driven architectural design and VLSI design realities. An innovative communication protocol minimizes power usage, and data transport costs among nodes were vastly reduced by eliminating the address bus, through local target address matching. Throughout the development process, the software and architecture teams were able to innovate alongside the circuit design team's implementation effort. A digital twin of the proposed hardware was developed early on to ensure that the technical implementation met the architectural specifications, and indeed the predicted performance metrics have now been thoroughly verified in real hardware test data. The resulting device is currently being used in a fielded edge sensor application; additional proofs of principle are in progress demonstrating the proof on the ground of this new real-world extremely low-power high-performance ASIC device., Comment: neural network accelerator, low-power design, instruction set design, parallel processors, digital twin

Extensive experiments suggest that motor coordination among human participants may contribute to social affinity and emotional attachment, which has great potential in the clinical treatment of social disorders or schizophrenia. Mirror game provides an effective experimental paradigm for studying social motor coordination. Nevertheless, the lack of movement richness prevents the emergence of high-level coordination in the existing one-dimensional experiments. To tackle this problem, this work develops a two-dimensional experimental paradigm of mirror game by playing waggle dance between two participants. In particular, an online control architecture of customized virtual player is created to coordinate with human player. Therein, an iterative learning control algorithm is proposed by integrating position tracking and behavior imitation with prescribed kinematic feature. Moreover, convergence analysis of control algorithm is conducted to guarantee the online performance of virtual player. Finally, the proposed control strategy is validated by matching experimental data and compared with other control methods using a set of performance indexes.

While next-token prediction is considered a promising path towards artificial general intelligence, it has struggled to excel in multimodal tasks, which are still dominated by diffusion models (e.g., Stable Diffusion) and compositional approaches (e.g., CLIP combined with LLMs). In this paper, we introduce Emu3, a new suite of state-of-the-art multimodal models trained solely with next-token prediction. By tokenizing images, text, and videos into a discrete space, we train a single transformer from scratch on a mixture of multimodal sequences. Emu3 outperforms several well-established task-specific models in both generation and perception tasks, surpassing flagship models such as SDXL and LLaVA-1.6, while eliminating the need for diffusion or compositional architectures. Emu3 is also capable of generating high-fidelity video via predicting the next token in a video sequence. We simplify complex multimodal model designs by converging on a singular focus: tokens, unlocking great potential for scaling both during training and inference. Our results demonstrate that next-token prediction is a promising path towards building general multimodal intelligence beyond language. We open-source key techniques and models to support further research in this direction., Comment: Project Page: https://emu.baai.ac.cn

Molecular dynamics (MD) is a powerful technique for studying microscopic phenomena, but its computational cost has driven significant interest in the development of deep learning-based surrogate models. We introduce generative modeling of molecular trajectories as a paradigm for learning flexible multi-task surrogate models of MD from data. By conditioning on appropriately chosen frames of the trajectory, we show such generative models can be adapted to diverse tasks such as forward simulation, transition path sampling, and trajectory upsampling. By alternatively conditioning on part of the molecular system and inpainting the rest, we also demonstrate the first steps towards dynamics-conditioned molecular design. We validate the full set of these capabilities on tetrapeptide simulations and show that our model can produce reasonable ensembles of protein monomers. Altogether, our work illustrates how generative modeling can unlock value from MD data towards diverse downstream tasks that are not straightforward to address with existing methods or even MD itself. Code is available at https://github.com/bjing2016/mdgen., Comment: NeurIPS 2024

Idioms represent a ubiquitous vehicle for conveying sentiments in the realm of everyday discourse, rendering the nuanced analysis of idiom sentiment crucial for a comprehensive understanding of emotional expression within real-world texts. Nevertheless, the existing corpora dedicated to idiom sentiment analysis considerably limit research in text sentiment analysis. In this paper, we propose an innovative approach to automatically expand the sentiment lexicon for idioms, leveraging the capabilities of large language models through the application of Chain-of-Thought prompting. To demonstrate the effectiveness of this approach, we integrate multiple existing resources and construct an emotional idiom lexicon expansion dataset (called EmoIdiomE), which encompasses a comprehensive repository of Chinese and English idioms. Then we designed the Dual Chain-of-Thoughts (DualCoTs) method, which combines insights from linguistics and psycholinguistics, to demonstrate the effectiveness of using large models to automatically expand the sentiment lexicon for idioms. Experiments show that DualCoTs is effective in idioms sentiment lexicon expansion in both Chinese and English. For reproducibility, we will release the data and code upon acceptance.

Large vision-language models (VLMs) have become state-of-the-art for many computer vision tasks, with in-context learning (ICL) as a popular adaptation strategy for new ones. But can VLMs learn novel concepts purely from visual demonstrations, or are they limited to adapting to the output format of ICL examples? We propose a new benchmark we call Spatial Visual Ambiguity Tasks (SVAT) that challenges state-of-the-art VLMs to learn new visuospatial tasks in-context. We find that VLMs fail to do this zero-shot, and sometimes continue to fail after finetuning. However, adding simpler data to the training by curriculum learning leads to improved ICL performance., Comment: 13 pages, 4 figures. Code released at https://github.com/groundlight/vlm-visual-demonstrations

Long-term memory is significant for agents, in which insights play a crucial role. However, the emergence of irrelevant insight and the lack of general insight can greatly undermine the effectiveness of insight. To solve this problem, in this paper, we introduce Multi-Scale Insight Agent (MSI-Agent), an embodied agent designed to improve LLMs' planning and decision-making ability by summarizing and utilizing insight effectively across different scales. MSI achieves this through the experience selector, insight generator, and insight selector. Leveraging a three-part pipeline, MSI can generate task-specific and high-level insight, store it in a database, and then use relevant insight from it to aid in decision-making. Our experiments show that MSI outperforms another insight strategy when planning by GPT3.5. Moreover, We delve into the strategies for selecting seed experience and insight, aiming to provide LLM with more useful and relevant insight for better decision-making. Our observations also indicate that MSI exhibits better robustness when facing domain-shifting scenarios.

We investigate the properties of domain walls arising from non-Abelian discrete symmetries, which we refer to as non-Abelian domain walls. We focus on $S_4$, one of the most commonly used groups in lepton flavour mixing models. The spontaneous breaking of $S_4$ leads to distinct vacua preserving a residual $Z_2$ or $Z_3$ symmetry. Five types of domain walls are found, labelled as SI, SII, TI, TII, and TIII, respectively, the former two separating $Z_2$ vacua and the latter three separating $Z_3$ vacua. We highlight that SI, TI and TIII may be unstable for some regions of the parameter space and decay to stable domain walls. Stable domain walls can collapse and release gravitational radiation for a suitable size of explicit symmetry breaking. A symmetry-breaking scale of order 100 TeV may explain the recent discovery of nanohertz gravitational waves by PTA experiments. For the first time, we investigate the properties of these domain walls, which we obtain numerically with semi-analytical formulas applied to compute the tension and thickness across a wide range of parameter space. We estimate the resulting gravitational wave spectrum and find that, thanks to their rich vacuum structure, non-Abelian domain walls manifest in a very interesting and complex phenomenology., Comment: 29 pages, 15 figures

Observations of the ultra-short period rocky exoplanet 55 Cancri e (55 Cnc e) indicate that the planet's dayside infrared radiation fluctuates by a factor of at least six on sub-weekly timescales, for unknown reasons. We propose a feedback mechanism where increased reflective clouds cool surface magma, subsequently reducing cloud formation, which may offer a potential explanation for these phenomena. In this mechanism, under less cloudy conditions, stellar radiation heats the surface magma, causing it to release more silicate vapor, which then condenses to form reflective clouds. Once formed, these clouds reduce stellar insolation at the surface, leading to surface cooling, which in turn reduces vapor supply, decreasing cloudiness. A time lag between the temperature increase of surface magma and the subsequent increase in cloudiness (likely due to lagged atmospheric transport of cloud-forming vapor) enables self-sustained oscillations in surface temperature and cloud reflectivity. These oscillations manifest as variations in both the emitted thermal radiation and the reflected stellar radiation, causing variability in secondary eclipse depths across different wavelengths without significantly affecting the transit depth. Using a simple model, we find that diverse planetary parameters can reproduce the observations. Additionally, we demonstrate that secondary eclipse depths at different wavelengths can oscillate out of phase, consistent with recent observations by the James Webb Space Telescope. Finally, we discuss observational strategies to test this proposed mechanism on 55 Cancri e. If confirmed, observable ocean-atmosphere dynamics on exoplanets would open a new window into the composition, evolution, and fate of rocky planet volatiles., Comment: Authors Luo and Loftus contributed equally. 17 pages, 3 figures, 2 tables. Associated code at https://github.com/kaitlyn-loftus/lava-limits