NeurIPS 2025 论文列表

Direct Preference Optimization (DPO) has emerged as an effective approach for aligning large language models (LLMs) with human preferences. However, its performance is highly dependent on the quality of the underlying human preference data. To address this bottleneck, prior work has explored various data selection strategies, but these methods often overlook the impact of the evolving states of the language model during the optimization process. In this paper, we introduce a novel problem: Sample Scheduling for DPO, which aims to dynamically and adaptively schedule training samples based on the model's evolving batch-wise states throughout preference optimization. To solve this problem, we propose SamS, an efficient and effective algorithm that adaptively selects samples in each training batch based on the LLM's learning feedback to maximize the potential generalization performance. Notably, without modifying the core DPO algorithm, simply integrating SamS significantly improves performance across tasks, with minimal additional computational overhead. This work points to a promising new direction for improving LLM alignment through batch-wise sample selection, with potential generalization to RLHF and broader supervised learning paradigms.

Recent advances in 3D scene reconstruction enable real-time viewing in virtual and augmented reality. To support interactive operations for better immersiveness, such as moving or editing objects, 3D scene inpainting methods are proposed to repair or complete the altered geometry. To support users in interacting (such as moving or editing objects) with the scene for the next level of immersiveness, 3D scene inpainting methods are developed to repair the altered geometry. However, current approaches rely on lengthy and computationally intensive optimization, making them impractical for real-time or online applications. We propose InstaInpaint, a reference-based feed-forward framework that produces 3D-scene inpainting from a 2D inpainting proposal within 0.4 seconds. We develop a self-supervised masked-finetuning strategy to enable training of our custom large reconstruction model (LRM) on the large-scale dataset. Through extensive experiments, we analyze and identify several key designs that improve generalization, textural consistency, and geometric correctness. InstaInpaint achieves a 1000$\times$ speed-up from prior methods while maintaining a state-of-the-art performance across two standard benchmarks. Moreover, we show that InstaInpaint generalizes well to flexible downstream applications such as object insertion and multi-region inpainting.

Continual model merging integrates independently fine-tuned models sequentially without access to the original training data, offering a scalable and efficient solution for continual learning. However, existing methods face two critical challenges: parameter interference among tasks, which leads to catastrophic forgetting, and limited adaptability to evolving test distributions. To address these issues, we introduce the task of Test-Time Continual Model Merging (TTCMM), which leverages a small set of unlabeled test samples during inference to alleviate parameter conflicts and handle distribution shifts. We propose MINGLE, a novel framework for TTCMM. MINGLE employs a mixture-of-experts architecture with parameter-efficient, low-rank experts, which enhances adaptability to evolving test distributions while dynamically merging models to mitigate conflicts. To further reduce forgetting, we propose Null-Space Constrained Gating, which restricts gating updates to subspaces orthogonal to prior task representations, thereby suppressing activations on old tasks and preserving past knowledge. We further introduce an Adaptive Relaxation Strategy that adjusts constraint strength dynamically based on interference signals observed during test-time adaptation, striking a balance between stability and adaptability. Extensive experiments on standard continual merging benchmarks demonstrate that MINGLE achieves robust generalization, significantly reduces forgetting, and consistently surpasses previous state-of-the-art methods by 7–9% on average across diverse task orders. Our code is available at: https://github.com/zihuanqiu/MINGLE

Prompt caching is critical for reducing latency and cost in LLM inference---OpenAI and Anthropic report up to 50–90\% cost savings through prompt reuse. Despite its widespread success, little is known about what constitutes an optimal prompt caching policy, particularly when optimizing tail latency—a metric of central importance to practitioners. The widely used Least Recently Used (LRU) policy can perform arbitrarily poor on this metric, as it is oblivious to the heterogeneity of conversation lengths. To address this gap, we propose Tail-Optimized LRU, a simple two-line modification that reallocates KV cache capacity to prioritize high-latency conversations by evicting cache entries that are unlikely to affect future turns. Though the implementation is simple, we prove its optimality under a natural stochastic model of conversation dynamics, providing the first theoretical justification for LRU in this setting---a result that may be of independent interest to the caching community. Experimentally, on real conversation data WildChat~\citep{zhao2024wildchat}, Tail-Optimized LRU achieves up to 27.5\% reduction in P90 tail Time to First Token latency and 23.9\% in P95 tail latency compared to LRU, along with up to 38.9\% decrease in SLO violations of 200ms. We believe this provides a practical and theoretically grounded option for practitioners seeking to optimize tail latency in real-world LLM deployments.

With the ever-increasing volume of visual data, the efficient and lossless transmission, along with its subsequent interpretation and understanding, has become a critical bottleneck in modern information systems. The emerged codebook-based solution utilize a globally shared codebook to quantize and dequantize each token, controlling the bpp by adjusting the number of tokens or the codebook size. However, for facial images—which are rich in attributes—such global codebook strategies overlook both the category-specific correlations within images and the semantic differences among tokens, resulting in suboptimal performance, especially at low bpp. Motivated by these observations, we propose a Switchable Token-Specific Codebook Quantization for face image compression, which learns distinct codebook groups for different image categories and assigns an independent codebook to each token. By recording the codebook group to which each token belongs with a small number of bits, our method can reduce the loss incurred when decreasing the size of each codebook group. This enables a larger total number of codebooks under a lower overall bpp, thereby enhancing the expressive capability and improving reconstruction performance. Owing to its generalizable design, our method can be integrated into any existing codebook-based representation learning approach and has demonstrated its effectiveness on face recognition datasets, achieving an average accuracy of 93.51\% for reconstructed images at 0.05 bpp.

This work proposes the Lattice Boltzmann Model (LBM) to learn real-world pixel dynamicity for visual tracking. LBM decomposes visual representations into dynamic pixel lattices and solves pixel motion states through collision-streaming processes. Specifically, the high-dimensional distribution of the target pixels is acquired through a multilayer predict-update network to estimate the pixel positions and visibility. The predict stage formulates lattice collisions among the spatial neighborhood of target pixels and develops lattice streaming within the temporal visual context. The update stage rectifies the pixel distributions with online visual representations. Compared with existing methods, LBM demonstrates practical applicability in an online and real-time manner, which can efficiently adapt to real-world visual tracking tasks. Comprehensive evaluations of real-world point tracking benchmarks such as TAP-Vid and RoboTAP validate LBM's efficiency. A general evaluation of large-scale open-world object tracking benchmarks such as TAO, BFT, and OVT-B further demonstrates LBM's real-world practicality.

Recent reasoning-focused language models achieve high accuracy by generating lengthy intermediate reasoning paths before producing final answers. While this approach is effective in solving problems that require logical thinking, long reasoning paths significantly increase memory usage and reduce throughput of token generation, limiting the practical deployment of such models. We propose Reasoning Path Compression (RPC), a training-free method that accelerates inference by leveraging the semantic sparsity of reasoning paths. RPC periodically compresses the KV cache by retaining cache entries that receive high importance score, which are computed using a selector window composed of recently generated queries. Experiments show that RPC improves generation throughput of QwQ-32B by up to 1.60$\times$ compared to the inference with full KV cache, with an accuracy drop of 1.2\% on the AIME 2024 benchmark. Our findings demonstrate that semantic sparsity in reasoning traces can be effectively exploited for compression, offering a practical path toward efficient deployment of reasoning LLMs. Our code is available at https://github.com/jiwonsong-dev/ReasoningPathCompression.

While tabular machine learning has achieved remarkable success, temporal distribution shifts pose significant challenges in real-world deployment, as the relationships between features and labels continuously evolve. Static models assume fixed mappings to ensure generalization, whereas adaptive models may overfit to transient patterns, creating a dilemma between robustness and adaptability. In this paper, we analyze key factors essential for constructing an effective dynamic mapping for temporal tabular data. We discover that evolving feature semantics—particularly objective and subjective meanings—introduce concept drift over time. Crucially, we identify that feature transformation strategies are able to mitigate discrepancies in feature representations across temporal stages. Motivated by these insights, we propose a feature-aware temporal modulation mechanism that conditions feature representations on temporal context, modulating statistical properties such as scale and skewness. By aligning feature semantics across time, our approach achieves a lightweight yet powerful adaptation, effectively balancing generalizability and adaptability. Benchmark evaluations validate the effectiveness of our method in handling temporal shifts in tabular data.

We study a two-sided market, wherein, price-sensitive heterogeneous customers and servers arrive and join their respective queues. A compatible customer-server pair can then be matched by the platform, at which point, they leave the system. Our objective is to design pricing and matching algorithms that maximize the platform's profit, while maintaining reasonable queue lengths. As the demand and supply curves governing the price-dependent arrival rates may not be known in practice, we design a novel online-learning-based pricing policy and establish its near-optimality. In particular, we prove a tradeoff among three performance metrics: $\tilde{O}(T^{1-\gamma})$ regret, $\tilde{O}(T^{\gamma/2})$ average queue length, and $\tilde{O}(T^{\gamma})$ maximum queue length for $\gamma \in (0, 1/6]$, significantly improving over existing results (Yang & Ying, 2024). Moreover, barring the permissible range of $\gamma$, we show that this trade-off between regret and average queue length is optimal up to logarithmic factors under a class of policies, matching the optimal one as in (Varma et al., 2023) which assumes the demand and supply curves to be known. Our proposed policy has two noteworthy features: a dynamic component that optimizes the tradeoff between low regret and small queue lengths; and a probabilistic component that resolves the tension between obtaining useful samples for fast learning and maintaining small queue lengths.

U‑shaped networks output logits at multiple spatial scales, each capturing a different blend of coarse context and fine detail. Yet, training still treats these logits in isolation—either supervising only the final, highest‑resolution logits or applying deep supervision with identical loss weights at every scale—without exploring *mixed‑scale* combinations. Consequently, the decoder output misses the complementary cues that arise only when coarse and fine predictions are fused. To address this issue, we introduce LoMix ($\underline{Lo}$gits $\underline{Mix}$ing), a Neural Architecture Search (NAS)‑inspired, differentiable plug-and-play module that **generates** new mixed‑scale outputs and **learns** how exactly each of them should guide the training process. More precisely, LoMix mixes the multi-scale decoder logits with four lightweight fusion operators: addition, multiplication, concatenation, and attention-based weighted fusion, yielding a rich set of synthetic “mutant’’ maps. Every original or mutant map is given a softplus loss weight that is co‑optimized with network parameters, mimicking a one‑step architecture search that automatically discovers the most useful scales, mixtures, and operators. Plugging LoMix into recent U-shaped architectures (i.e., PVT‑V2‑B2 backbone with EMCAD decoder) on Synapse 8‑organ dataset improves DICE by +4.2\% over single‑output supervision, +2.2\% over deep supervision, and +1.5\% over equally weighted additive fusion, all with **zero** inference overhead. When training data are scarce (e.g., one or two labeled scans, 5\% of the trainset), the advantage grows to +9.23\%, underscoring LoMix’s data efficiency. Across four benchmarks and diverse U-shaped networks, LoMiX improves DICE by up to +13.5\% over single-output supervision, confirming that learnable weighted mixed‑scale fusion generalizes broadly while remaining data efficient, fully interpretable, and overhead-free at inference. Our implementation is available at https://github.com/SLDGroup/LoMix.

Accurate video forecasting enables autonomous vehicles to anticipate hazards, robotics and surveillance systems to predict human intent, and environmental models to issue timely warnings for extreme weather events. However, existing methods remain limited: transformers rely on global attention with quadratic complexity, making them impractical for high-resolution, long-horizon video prediction, while convolutional and recurrent networks suffer from short-range receptive fields and vanishing gradients, losing key information over extended sequences. To overcome these challenges, we introduce VideoTitans, the first architecture to adapt the gradient-driven Titans memory—originally designed for language modelling to video prediction. VideoTitans integrates three core ideas: (i) a sliding-window attention core that scales linearly with sequence length and spatial resolution, (ii) an episodic memory that dynamically retains only informative tokens based on a gradient-based surprise signal, and (iii) a small set of persistent tokens encoding task-specific priors that stabilize training and enhance generalization. Extensive experiments on Moving-MNIST, Human3.6M, TrafficBJ and WeatherBench benchmarks show that VideoTitans consistently reduces computation (FLOPs) and achieves competitive visual fidelity compared to state-of-the-art recurrent, convolutional, and efficient-transformer methods. Comprehensive ablations confirm that each proposed component contributes significantly.

As large language models (LLMs) become increasingly integrated into applications serving users across diverse cultures, communities, and demographics, it is critical to align LLMs with pluralistic human values beyond average principles (e.g., HHH). In psychological and social value theories such as Schwartz’s Value Theory, pluralistic values are represented by multiple value dimensions paired with various priorities. However, existing methods encounter two challenges when aligning with such fine-grained value objectives: 1) they often treat multiple values as independent and equally important, ignoring their interdependence and relative priorities (value complexity); 2) they struggle to precisely control nuanced value priorities, especially those underrepresented ones (value steerability). To handle these challenges, we propose COUPLE, a COUnterfactual reasoning framework for PLuralistic valuE alignment. It introduces a structural causal model (SCM) to feature complex interdependency and prioritization among features, as well as the causal relationship between high-level value dimensions and behaviors. Moreover, it applies counterfactual reasoning to generate outputs aligned with any desired value objectives. Benefitting from explicit causal modeling, COUPLE also provides better interpretability. We evaluate COUPLE on two datasets with different value systems and demonstrate that COUPLE advances other baselines across diverse types of value objectives. Our code is available at https://github.com/microsoft/COUPLE.

Floorplanning is the initial step in the physical design process of Electronic Design Automation (EDA), directly influencing subsequent placement, routing, and final power of the chip. However, the solution space in floorplanning is vast, and current algorithms often struggle to explore it sufficiently, making them prone to getting trapped in local optima. To achieve efficient floorplanning, we propose **CORE**, a general and effective solution optimization framework that synergizes Evolutionary Algorithms (EAs) and Reinforcement Learning (RL) for high-quality layout search and optimization. Specifically, we propose the Clustering-based Diversified Evolutionary Search that directly perturbs layouts and evolves them based on novelty and performance. Additionally, we model the floorplanning problem as a sequential decision problem with B*-Tree representation and employ RL for efficient learning. To efficiently coordinate EAs and RL, we propose the reinforcement-driven mechanism and evolution-guided mechanism. The former accelerates population evolution through RL, while the latter guides RL learning through EAs. The experimental results on the MCNC and GSRC benchmarks demonstrate that CORE outperforms other strong baselines in terms of wirelength and area utilization metrics, achieving a 12.9\% improvement in wirelength. CORE represents the first evolutionary reinforcement learning (ERL) algorithm for floorplanning, surpassing existing RL-based methods. The code is available at https://github.com/yeshenpy/CORE.

Data integration approaches are increasingly used to enhance the efficiency and generalizability of studies. However, a key limitation of these methods is the assumption that outcome measures are identical across datasets -- an assumption that often does not hold in practice. Consider the following opioid use disorder (OUD) studies: the XBOT trial and the POAT study, both evaluating the effect of medications for OUD on withdrawal symptom severity (not the primary outcome of either trial). While XBOT measures withdrawal severity using the subjective opiate withdrawal scale, POAT uses the clinical opiate withdrawal scale. We analyze this realistic yet challenging setting where outcome measures differ across studies and where neither study records both types of outcomes. Our paper studies whether and when integrating studies with disparate outcome measures leads to efficiency gains. We introduce three sets of assumptions -- with varying degrees of strength -- linking both outcome measures. Our theoretical and empirical results highlight a cautionary tale: integration can improve asymptotic efficiency only under the strongest assumption linking the outcomes. However, misspecification of this assumption leads to bias. In contrast, a milder assumption may yield finite-sample efficiency gains, yet these benefits diminish as sample size increases. We illustrate these trade-offs via a case study integrating the XBOT and POAT datasets to estimate the comparative effect of two medications for opioid use disorder on withdrawal symptoms. By systematically varying the assumptions linking the SOW and COW scales, we show potential efficiency gains and the risks of bias. Our findings emphasize the need for careful assumption selection when fusing datasets with differing outcome measures, offering guidance for researchers navigating this common challenge in modern data integration.

Proprietary large language models (LLMs) exhibit strong generalization capabilities across diverse tasks and are increasingly deployed on edge devices for efficiency and privacy reasons. However, deploying proprietary LLMs at the edge without adequate protection introduces critical security threats. Attackers can extract model weights and architectures, enabling unauthorized copying and misuse. Even when protective measures prevent full extraction of model weights, attackers may still perform advanced attacks, such as fine-tuning, to further exploit the model. Existing defenses against these threats typically incur significant computational and communication overhead, making them impractical for edge deployment. To safeguard the edge-deployed LLMs, we introduce CoreGuard, a computation- and communication-efficient protection method. CoreGuard employs an efficient protection protocol to reduce computational overhead and minimize communication overhead via a propagation protocol. Extensive experiments show that CoreGuard achieves upper-bound security protection with negligible overhead.

Audio captioning systems face a fundamental challenge: teacher-forcing training creates exposure bias that leads to caption degeneration during inference. While contrastive methods have been proposed as solutions, they typically fail to capture the crucial temporal relationships between acoustic and linguistic modalities. We address this limitation by introducing the unbiased sliced Wasserstein RBF (USW-RBF) kernel with rotary positional embedding, specifically designed to preserve temporal information across modalities. Our approach offers a practical advantage: the kernel enables efficient stochastic gradient optimization, making it computationally feasible for real-world applications. Building on this foundation, we develop a complete audio captioning framework that integrates stochastic decoding to further mitigate caption degeneration. Extensive experiments on AudioCaps and Clotho datasets demonstrate that our method significantly improves caption quality, lexical diversity, and text-to-audio retrieval accuracy. Furthermore, we demonstrate the generalizability of our USW-RBF kernel by applying it to audio reasoning tasks, where it enhances the reasoning capabilities of large audio language models on the CompA-R in terms of correctness and quality. Our kernel also improves the reasoning accuracy of the MMAU-test-mini benchmarks by $4\%$. These results establish our approach as a powerful and generalizable solution for cross-modal alignment challenges in audio-language tasks.

Offline reinforcement learning (RL) aims to learn a policy from a fixed dataset without additional environment interaction. However, effective offline policy learning often requires a large and diverse dataset to mitigate epistemic uncertainty. Collecting such data demands substantial online interactions, which are costly or infeasible in many real-world domains. Therefore, improving policy learning from limited offline data—achieving high data efficiency—is critical for practical offline RL. In this paper, we propose a simple yet effective plug-and-play pretraining framework that initializes the feature representation of a $Q$-network to enhance data efficiency in offline RL. Our approach employs a shared $Q$-network architecture trained in two stages: pretraining a backbone feature extractor with a transition prediction head; training a $Q$-network—combining the backbone feature extractor and a $Q$-value head—with *any* offline RL objective. Extensive experiments on the D4RL, Robomimic, V-D4RL, and ExoRL benchmarks show that our method substantially improves both performance and data efficiency across diverse datasets and domains. Remarkably, with only **10\%** of the dataset, our approach outperforms standard offline RL baselines trained on the full data.

Generating hyperspectral images (HSIs) from RGB images through spectral reconstruction can significantly reduce the cost of HSI acquisition. In this paper, we propose a Fractal-Based Recursive Spectral Reconstruction Network (FRN), which differs from existing paradigms that attempt to directly integrate the full-spectrum information from the R, G, and B channels in a one-shot manner. Instead, it treats spectral reconstruction as a progressive process, predicting from broad to narrow bands or employing a coarse-to-fine approach for predicting the next wavelength. Inspired by fractals in mathematics, FRN establishes a novel spectral reconstruction paradigm by recursively invoking an atomic reconstruction module. In each invocation, only the spectral information from neighboring bands is used to provide clues for the generation of the image at the next wavelength, which follows the low-rank property of spectral data. Moreover, we design a band-aware state space model that employs a pixel-differentiated scanning strategy at different stages of the generation process, further suppressing interference from low-correlation regions caused by reflectance differences. Through extensive experimentation across different datasets, FRN achieves superior reconstruction performance compared to state-of-the-art methods. Code is available at https://github.com/mongko007/frn.

Optimization modeling is fundamental to decision-making in fields such as supply chain management, logistics, and financial engineering, but its complexity presents a major barrier to adoption. Automating model creation from natural language is key to improving efficiency and access. However, while Large Language Models (LLMs) are a promising tool for this, they often produce flawed or infeasible results due to errors and hallucinations. To address this issue, we propose Solver-Informed Reinforcement Learning (SIRL), a framework that uses Reinforcement Learning with Verifiable Reward to improve LLMs’ ability to generate accurate and executable optimization models. Specifically, SIRL automatically assesses the executable code and the instance-level mathematical model represented by the associated .lp files. This process yields precise feedback on syntactic validity, feasibility, and solution quality, which serves as a direct reward signal to guide the reinforcement learning process. Furthermore, this verification mechanism also supports our instance-enhanced self-consistency method for creating high-quality training data. Extensive experiments on diverse public benchmarks demonstrate that models trained with our SIRL framework achieve state-of-the-art performance, substantially outperforming existing methods in generating accurate and executable optimization models. Specifically, our SIRL-32B model surpasses DeepSeek-V3 and OpenAI-o3 on the majority of these benchmarks. Our code is publicly available at https://github.com/Cardinal-Operations/SIRL.