论文检索

Deep Learning · Large Language Models

Aneesh Pappu, Batu El, Hancheng Cao, Carmelo di Nolfo, Yanchao Sun, Meng Cao, James Zou

Multi-agent LLM systems are increasingly deployed as autonomous collaborators, where agents interact freely rather than execute fixed, pre-specified workflows. In such settings, effective coordination cannot be fully designed in advance and must instead emerge through interaction. However, most prior work enforces coordination through fixed roles, workflows, or aggregation rules, leaving open the question of how well self-organizing teams perform when coordination is unconstrained. Drawing on organizational psychology, we study whether self-organizing LLM teams achieve *strong synergy*, where team performance matches or exceeds the best individual member. Across human-inspired and frontier ML benchmarks, we find that---unlike human teams---LLM teams consistently fail to match their expert agent's performance, even when explicitly told who the expert is, incurring performance losses of up to 37.6%. Decomposing this failure, we show that expert leveraging, rather than identification, is the primary bottleneck. Conversational analysis reveals a tendency toward integrative compromise---averaging expert and non-expert views rather than appropriately weighting expertise---which increases with team size and correlates negatively with performance. Interestingly, this consensus-seeking behavior improves robustness to adversarial agents, suggesting a trade-off between alignment and effective expertise utilization. Our findings reveal a significant gap in the ability of self-organizing multi-agent teams to harness the collective expertise of their members.

Theory · Optimization

Xiaoyuan Liang, Sebastian Loeschcke, Mads Toftrup, Anima Anandkumar

Low-precision formats such as BF16 and FP8 can greatly improve training efficiency, but fully low-precision training often degrades accuracy under standard optimizers. We identify a key cause: additive updates can vanish under coarse mantissa resolution. We introduce M+Adam, an optimizer for stable low-precision training that operates on a mantissa--exponent decomposition of weights and carries out Adam-Madam updates in parallel. Madam is a multiplicative analogue of Adam, where instead of applying additive updates, it performs updates multiplicatively, which is naturally suited for updating exponents. Building on this idea, \method applies additive updates to the mantissa and multiplicative updates to the exponent in parallel. We demonstrate the complementary failure modes of purely additive or multiplicative updates under quantization and thus, our method that combines both can overcome all these failure modes. We establish a monotone descent guarantee under standard smoothness assumptions for our method. Under a challenging setting where both the weight and compute are in FP8, M+Adam substantially outperforms AdamW (e.g., by 10.51\% perplexity at 350M). Moreover, M+Adam enables stable BF16 training without stochastic rounding and consistently outperforms AdamW across 60M-350M models and $1$-$8\times$ Chinchilla budgets.

Optimization · Everything Else

Fei Liu, Rui Zhang, Shunyu Yao, Qinglong Hu, kefeng zheng, Zhichao Lu, Qingfu Zhang

Designing heuristic algorithms for complex optimization problems is a time-consuming and expert-driven process. Recently, Automated Heuristic Design (AHD) using Large Language Models (LLMs) has shown significant promise for automating algorithm development. However, existing works mainly rely on programs to represent heuristics, which are inherently taskspecific and fail to generalize as effectively as established metaheuristics like tabu search or guided local search. To bridge this gap, we introduce Multi-Task Hierarchical Search (MTHS), an LLM-guided evolutionary method that co-designs general-purpose metaheuristics and task-specific programs. MTHS employs a hierarchical representation and adopts a two-level evolution framework to evolve task-agnostic metaheuristics and task-specific program implementations simultaneously across multiple heuristic design tasks. During this evolution, a knowledge transfer mechanism allows learning from elite programs designed for other tasks. We evaluated MTHS on distinct combinatorial optimization problems, where it outperforms both commonly-used heuristics and existing LLM-driven AHD approaches. Our results demonstrate that the hierarchical representations facilitate effective multi-task AHD, and the evolved metaheuristics exhibit strong generalization to related tasks.

Theory · Online Learning and Bandits

Heesang Ann, Hyun-jun Choi, Taehyun Hwang, Younghoon Shin, Haeju Cheong, Min-hwan Oh

We study generalized linear bandits with memory, a non-stationary setting in which rewards depend on past actions through a finite memory matrix. Building on prior work for linear models Clerici et al.,(2024), we show that the previously known $\tilde{\mathcal{O}}(T^{3/4})$ regret stems from a loose analysis based on cyclic proxy policies, and we refine the analysis to recover a $\tilde{\mathcal{O}}(\sqrt{T})$ regret rate in the linear case. We then extend this improvement to generalized linear models and propose a block-wise algorithm based on shrinkage-based confidence bounds. Our algorithm achieves a regret bound of $\tilde{\mathcal{O}}(\sqrt{mT}+ d\sqrt{T}+\sqrt{\kappa} d^{2} m^{1/4} T^{1/4} + \kappa d^{2})$, where $d$ denotes the feature dimension, $m$ the memory length, and $\kappa$ a curvature parameter of the link function, thereby attaining a $\sqrt{T}$ rate despite nonlinear rewards and memory effects. To the best of our knowledge, this analysis provides a unified treatment of memory-induced non-stationarity and nonlinear link functions, while ensuring that the leading regret term is independent of the curvature of the link function. We conduct numerical experiments that are consistent with our theoretical findings.

Theory · Online Learning and Bandits

Wooseong Cho, Ji Hyeong Park, Min-hwan Oh

We study bilinear bandits with partially observable features on both the user and item sides. In each of $T$ rounds, the learner selects an arm and observes only the reward for the chosen pair. The reward model is linear in the user and item features with an unknown parameter matrix. Existing literature commonly reduces this problem to a linear bandit via a Kronecker product representation of user and item features, at the cost of increased dimensionality. We propose \texttt{BiRoLF}, an algorithm robust to latent features, which directly leverages the bilinear structure without such linearization. It enhances feature selection by augmenting the null space of the observed features and employs doubly robust (DR) estimation to impute unobserved rewards for unselected arms, constructing unbiased pseudo-rewards. We estimate the parameters using Lasso regularization, which promotes sparsity in the coefficients of latent components orthogonal to the observed features. \texttt{BiRoLF} achieves a $\tilde{O}(\sqrt{(d_x + d_{h_x})(d_y + d_{h_y}) T})$ regret bound, where $d_x$ and $d_y$ are the dimensions of the observable feature vectors, $d_{h_x}$ and $d_{h_y}$ denote the numbers of nonzero coefficients in the components orthogonal to the observed features. We segment cases by the relationship between observable and latent features and find that \texttt{BiRoLF} achieves strong regret performance while outperforming competing methods in computational metrics, reducing the overhead of feature linearization.

Reinforcement Learning · Batch/Offline

Hyungkyu Kang, Byeongchan Kim, Min-hwan Oh

Offline goal-conditioned reinforcement learning (GCRL) provides a practical framework for obtaining goal-reaching policies from fixed datasets. However, learning a reliable goal-conditioned value function in long-horizon tasks remains challenging. In this paper, we identify erroneous generalization in goal-conditioned value functions as a fundamental bottleneck, and demonstrate that appropriate inductive bias in the value function is crucial for addressing the bottleneck. Building on these findings, we propose Latent-Aligned Value Learning (LAVL), an offline GCRL algorithm that integrates latent-representation-based value generalization with hierarchical planning in a unified framework. Extensive numerical experiments on OGBench demonstrate that LAVL consistently outperforms existing offline GCRL methods, achieving the highest performance on **20** out of 22 datasets. Notably, LAVL exhibits strong performance in long-horizon tasks and trajectory stitching datasets, where prior methods suffer significant performance degradation.

Deep Learning · Large Language Models

Hao Li, Hao Wan, Yixue Huang, Yuzhou Chen, Yulia Gel, Hao Jiang

Large language models (LLMs) are increasingly used in scientific discovery, system modeling, and decision-making, prompting interest in their ability to reason over complex structured data. Existing benchmarks primarily focus on static or local graph reasoning, overlooking the high-order structures in real-world systems whose global properties evolve across multiple scales. We introduce LLM4PH, a benchmark that evaluates multi-scale structural reasoning through the lens of persistent homology (PH), a topological framework for tracking structural evolution. LLM4PH decomposes the PH pipeline into interpretable reasoning tasks spanning synthetic and real-world graphs, revealing that most models struggle with reasoning over structural transitions and persistence. Beyond task-level evaluation, we perform cross-task ablations on prompt encoding and transfer, explore post-training effects, and construct a compositional PH pipeline to assess end-to-end performance. Our results provide the first in-depth view of how well LLMs bridge discrete graph structures with continuous topological abstraction, and offer insights into their potential for structure-aware scientific reasoning.

Deep Learning · Large Language Models

Yixing Xu, Guanchen Li, Chao Li, Xuanwu Yin, Dong Li, Spandan Tiwari, Ashish Sirasao, Emad Barsoum

Low-rank adaptation (LoRA) is one of the most widely used parameter-efficient fine-tuning (PEFT) methods for adapting pre-trained large language models (LLMs) to downstream tasks. Although LoRA significantly reduces the number of trainable parameters and lowers fine-tuning costs, its performance is often limited by the inherent low-rank assumption. In this paper, we revisit the notion of rank for LoRA update matrices and show that the standard matrix rank fails to capture duplicated directions and redundancy in the update subspace. Motivated by this analysis, we argue that the Kruskal rank offers a more informative criterion for characterizing update diversity. We therefore propose **Low Kruskal Rank Adaptation** (LoKRA), a new PEFT algorithm with provable theoretical guarantees that mitigates the limitations of LoRA. We further introduce LoKRA$^+$, an enhanced variant that provides a tighter theoretical lower bound on the Kruskal rank and yields stronger empirical performance. Experiments on multiple LLMs show that our approach consistently outperforms LoRA and other baselines, establishing state-of-the-art performance across a range of benchmarks.

Deep Learning · Large Language Models

Xiaopu Wang, Zelin He, Chengyuan Liu, Runze Li

Logit-based watermarking is a widely used mechanism for identifying LLM generated content, yet its effectiveness is governed by a fundamental trade-off between detectability and semantic distortion. Existing analyses provide limited guidance for principled hyperparameter selection, leaving practical deployments reliant on heuristic tuning. In this work, we develop a power-calibrated statistical framework that establishes explicit quantitative relationships between watermark hyperparameters, detection power, and distortion. This characterization transforms watermark design into a guided optimization problem. Building on these results, we derive practical parameter selection procedures that achieve optimal trade-offs under constraints. Extensive experiments across multiple language models and datasets validate the theory and demonstrate that the proposed framework consistently identifies Pareto-optimal points.

Theory · Online Learning and Bandits

Chaiwon Kim, Jongyeong Lee, Min-hwan Oh

We study the decoupled multi-armed bandit problem, where the learner selects one arm for exploration and one arm for exploitation separately at each round. In this setting, the loss of the explored arm is observed but not incurred, whereas the loss of the exploited arm is incurred without being observed. We propose an efficient Follow-the-Perturbed-Leader (FTPL) policy that achieves Best-of-Both-Worlds (BOBW) guarantee with constant regret in the stochastic regime and optimal $\mathcal{O}(\sqrt{KT})$ regret in the adversarial regime. A key feature of our method is that it completely avoids both the convex optimization required by prior BOBW policy, and the resampling procedures that are typically used in FTPL bandit policies. This allows FTPL to fully realize its computational efficiency advantages, and thus leads to substantial reductions in computational cost. We empirically confirm that our policy not only improves the runtime but also demonstrates superior regret performance in both regimes.

Social Aspects · Safety

Maciej Chrabaszcz, Filip Szatkowski, Bartosz Wójcik, Jan Dubiński, Tomasz Trzcinski, Sebastian Cygert

Although modern LLMs are aligned with human values during post-training, robust moderation remains essential to prevent harmful outputs at deployment time. Existing approaches suffer from performance-efficiency trade-offs and are difficult to customize to user-specific requirements. Motivated by this gap, we introduce Multi-Layer Prototype Moderator (MLPM), a lightweight and highly customizable input moderation tool. We propose leveraging prototypes of intermediate representations across multiple layers to improve moderation quality while maintaining high efficiency. By design, our method adds negligible overhead to the generation pipeline and can be seamlessly applied to any model. MLPM achieves state-of-the-art performance on diverse moderation benchmarks and demonstrates strong scalability across model families of various sizes. Moreover, we show that it integrates smoothly into end-to-end moderation pipelines and further improves response safety when combined with output moderation techniques. Overall, our work provides a practical and adaptable solution for safe, robust, and efficient LLM deployment.

Theory · Online Learning and Bandits

Joongkyu Lee, Min-hwan Oh

We study optimal experimental design for multinomial logit (MNL) bandits, where an agent repeatedly selects a subset of $K$ items from a ground set of size $N$ and observes single-choice feedback. Unlike linear or generalized linear bandits, MNL bandits have a combinatorial action space, which makes classical optimal design approaches and naive optimization over all subsets computationally intractable. We propose a computationally efficient optimal design framework for MNL models that achieves both statistical efficiency and scalability through two complementary approaches: (i) an exact or certified-approximate reformulation of the design oracle as a $0$-$1$ mixed-integer linear program (MILP) with solver-certified early stopping, and (ii) a fully polynomial-time lifted design that replaces the nonlinear objective with a tractable surrogate. Using the Kiefer-Wolfowitz equivalence theorem, we establish near G-optimality guarantees and characterize the induced statistical-computational trade-offs. As an application, we develop a best assortment identification algorithm for MNL bandits with linear utilities and non-uniform revenues, and prove an instance-dependent sample complexity of $\tilde{\mathcal{O}}\big(\frac{d \log N}{\Delta^2}\big)$, where $d$ is the feature dimension, $N$ is the number of arms, and $\Delta$ is the minimum revenue gap.

Deep Learning · Large Language Models

Rizhen Hu, Yuan Cao, Boao Kong, Mou Sun, Kun Yuan

Sparse Mixture-of-Experts (MoE) models scale Transformers efficiently but suffer from expert overlap, where different experts process similar tokens and learn redundant functions, resulting in ambiguous routing and underutilized capacity. While architectural solutions like DeepSeek-style shared experts promote specialization, they require substantial structural modifications and rely solely on intra-layer signals. We propose two plug-and-play auxiliary losses that enhance MoE specialization and routing efficiency without modifying routers or model architectures. First, an intra-layer specialization loss penalizes cosine similarity between experts' SwiGLU activations on identical tokens, encouraging experts to specialize in complementary functions. Second, a cross-layer dependency loss maximizes joint Top-$k$ routing probabilities across adjacent layers, establishing coherent expert pathways through network depth while reinforcing intra-layer specialization. Both losses are orthogonal to the standard load-balancing loss and compatible with shared-expert and vanilla Top-$k$ MoE architectures. We implement both losses as a drop-in Megatron-LM module. Extensive experiments across pre-training, fine-tuning, and zero-shot benchmarks demonstrate consistent task gains, higher expert specialization, and lower-entropy routing; together, these improvements translate into faster inference via more stable expert pathways.

Probabilistic Methods · Everything Else

Mengqi Chen, Thomas Berrett, Theodoros Damoulas, Michele Caprio

Distributionally robust optimisation (DRO) minimises the worst-case expected loss over an ambiguity set that can capture distributional shifts in out-of-sample environments. While Huber (linear-vacuous) contamination is a classical minimal-assumption model for an $\varepsilon$-fraction of arbitrary perturbations, including it in an ambiguity set can make the worst-case risk infinite and the DRO objective vacuous unless one imposes strong boundedness or support assumptions. We address these challenges by introducing bulk-calibrated credal ambiguity sets: we learn a high-mass bulk set from data while considering contamination inside the bulk and bounding the remaining tail contribution separately. This leads to a closed-form, finite $\mathrm{mean}+\sup$ robust objective and tractable linear or second-order cone programs for common losses and bulk geometries. Through this framework, we highlight and exploit the equivalence between the imprecise probability (IP) notion of upper expectation and the worst-case risk, demonstrating how IP credal sets translate into DRO objectives with interpretable tolerance levels. Experiments on heavy-tailed inventory control, geographically shifted house-price regression, and demographically shifted text classification show competitive robustness-accuracy trade-offs and efficient optimisation times, using Bayesian, frequentist, or empirical reference distributions.

Social Aspects · Security

Zhenxiong Yan, Suhang Yao, Yu Liu, Wenqiang Jin

Large language models (LLMs) are trained at significant computational and data cost, making them valuable intellectual property (IP). Existing IP verification methods primarily rely either on invasive watermarking that degrades model utility, or on superficial behavioral signatures disrupted by fine-tuning and model merging. This apparent trade-off between model utility and IP protection has constrained practical deployment. We challenge this trade-off and propose CircuitPrint, a non-invasive IP fingerprinting framework that enables robust verification through standard model queries by leveraging stable internal computational circuits of LLMs. We show that these circuits function as a persistent computational backbone across model derivatives, allowing them to serve as stable fingerprints for distinguishing LLMs. Building on this stability, CircuitPrint constructs IP signatures by identifying mechanistically essential supernodes that causally produce specific predictions within these circuits. Specifically, trigger queries are synthesized to replicate the internal suppression of these supernodes, thereby inducing distinctive and observable output shifts. Experimental results demonstrate that CircuitPrint substantially outperforms existing baselines while remaining robust under aggressive fine-tuning and model merging, effectively resolving this trade-off without altering model parameters.

General Machine Learning · Data

Chenyang Shao, Fengli Xu, Yong Li

AI agents have seen widespread adoption in information retrieval for scientific research, giving rise to tools such as *Deep Research*. However, existing retrieval agents mainly rely on keyword- or embedding-based methods. While effective at capturing content-level similarities, they struggle to understand complex relational networks among scientific papers, such as identifying corroborating or conflicting studies and tracing technological lineages. This fundamental limitation often results in fragmented knowledge structures, misinterpreted research sentiment, and ineffective modeling of collective scientific progress. To address this limitation, we introduce **SciNet**, the first **Sci**entific **Net**work relation-aware dataset for information retrieval agents. Built on a meta-database of 269 million papers across 7 disciplines and containing 8,940 carefully designed tasks, SciNet systematically captures three levels of relational understanding: ego-centric retrieval of papers with novel knowledge structures, pairwise identification of scholarly relationships, and path-wise reconstruction of scientific evolution. Extensive evaluation of three categories of retrieval agents shows that their accuracy on relation-aware tasks often falls below 20%, highlighting a fundamental shortcoming of current retrieval paradigms. Importantly, in a downstream literature review application, agents empowered with SciNet achieve a 25.3% improvement in review quality, highlighting the critical value of relation-aware retrieval for deepening scientific insights. We publicly release SciNet at [https://anonymous.4open.science/r/SciNet/](https://anonymous.4open.science/r/SciNet/) to support future research.

Applications · Robotics

Guanren Qiao, Ruixiang Ouyang, Sheng Xu, Ruixing Jin, Yueci Deng, Yunxin Tai, Kui Jia, Guiliang Liu

Real-World Reinforcement Learning (RL) has shown significant potential in robotic manipulation tasks. However, many methods still require substantial human-in-the-loop involvement to complete contact-rich tasks, especially when there are disruptions such as visual backgrounds or positional changes. To address this, we propose the Focus Then Contact (FTC), a lightweight and low-cost method to accelerate the convergence of human-in-the-loop real-world RL for contact-rich tasks. FTC leverages residual RL to provide base actions, helping the system quickly reach the target regions and improve sample efficiency. Additionally, FTC integrates an affordance-guided reward that drives the real-world RL system to quickly focus on key regions of interest, making it possible for the robotic arm to continuously engage with these goal areas through force-control feedback. At the same time, we optimize the human-in-the-loop implementation to prevent conflicts with RL over control of the robotic arm. We demonstrate the effectiveness of FTC on 6 contact-rich tasks, where it outperforms baseline methods in achieving high success rates and speeds up robotic contact-rich task learning under a real-world RL setting. Video materials can be seen in \url{https://anonymous.4open.science/api/repo/FTC-website-BB5E/file/index.html}.

Probabilistic Methods · Everything Else

Yingyan Zeng, Zipan Huang, Xiaoyu Chen

Existing online change-point detection (CPD) methods rely on fixed-dimensional Euclidean summaries, implicitly assuming that distributional changes are well captured by moment-based or feature-based representations. They can obscure important changes in distributional shape or geometry. We propose a geometry-aware CPD framework that treats streaming batch data as a stochastic process on the 2-Wasserstein space. Our method detects changes in the law of this process by mapping each empirical distribution to a tangent space relative to a pre-change Fréchet barycenter, yielding a reference-centered local linearization of 2-Wasserstein space. This representation enables sequential detectors by adapting classical multivariate monitoring statistics to tangent fields. We provide theoretical guarantees and demonstrate, via synthetic and real-world experiments, that our approach detects complex distributional shifts with reduced detection delay at matched $\mathrm{ARL}_0$ compared with moments-based and model-free baselines.

Deep Learning · Large Language Models

Siyuan Liu, Tinghong Chen, Xinghan Li, Yifei Wang, Jingzhao Zhang

Data selection during supervised fine-tuning (SFT) can critically change the behavior of large language models (LLMs). Although existing work has studied the effect of selecting data based on heuristics such as perplexity, difficulty, or length, the reported findings are often inconsistent or context-dependent. In this work, we systematically study the role of data difficulty in fine-tuning from both empirical and theoretical perspectives, and find that there is no universally optimal difficulty level; rather, its effectiveness depends on the dataset size. We show that for a fixed data budget, there exists an optimal data difficulty for SFT, and that this optimal difficulty shifts toward harder data as the data budget increases. To explain this phenomenon, we conduct controlled synthetic experiments that reveal a simple underlying mechanism: the interplay between the (in-distribution) generalization gap and the extrapolation gap. We further support this mechanism through a theoretical analysis using PAC-Bayesian generalization bounds. Overall, our results clarify how data size and difficulty jointly affect the trade-off between generalization and extrapolation in SFT, providing guidance for difficulty-based data selection under certain model and data conditions.

Applications · Language, Speech and Dialog

Chenyang Shao, Jiahe Liu, Fengli Xu, Yong Li

Scientific illustration figures are essential for depicting research works' conceptual designs, methodology, and experimental workflows, playing a pivotal role in communicating complex academic insights. However, creating high-quality scientific illustrations remains a labor-intensive task for human scientists. While recent generative image models have advanced prompt-based editing, the synthesis of fully **editable** figures remains a fundamental challenge. Valid editability involves structured transformations of graphical elements, scales, attributes, and text, rather than simple pixel-level changes. Existing models generate raster outputs that do not support manual correction or layout adjustment, limiting their utility in scientific publishing, where editable vector figures are typically required for submission. To address this challenge, we introduce **LiveFigure**, an agentic framework driven by VLM agents that imitates the multi-step drawing workflow of human researchers. It first plans figure blueprints by drawing inspiration from high-quality references in previous works, then generates executable scripts that produce figures via the PowerPoint interface based on skills and experience, and finally refines the outputs with targeted visual diagnostics, producing fully vectorized, editable figures that meet publication standards. Extensive experiments demonstrate that LiveFigure generates inherently editable figures that are both visually clear and aesthetically appealing, achieving 80% publication-readiness within just 17 manual edits—far surpassing the 24% rate of the strongest baseline, NanoBanna. Human preference studies further validate this advantage, with LiveFigure securing a 60% win rate against NanoBanna. Our code is available at [https://anonymous.4open.science/r/LiveFigure](https://anonymous.4open.science/r/LiveFigure).