LLM-based agents already operate in production across many industries, yet we lack a clear understanding of which technical methods make these deployments successful. We present the first systematic study of Characterizing Agents in Production (CAP) using first-hand data from agent developers. We conducted 20 in-depth case studies through interviews and surveyed 306 practitioners across 26 domains. We examine why organizations build agents, how they build them, how they evaluate them, and the key challenges they face in deployment. Our findings show that production agents rely on simple, controllable approaches: 68% execute at most 10 steps before human intervention, 70% rely on prompting off-the-shelf models rather than weight tuning, and 74% depend primarily on human evaluation. Reliability—defined as consistent correct behavior over time—emerges as the dominant challenge, which practitioners address through system-level design choices. CAP documents the current state of production agents, providing the research community with visibility into real-world deployment practices and underexplored research opportunities.
论文检索
输入标题、作者或关键词，从 13,033 篇学术成果中精准定位
Hierarchical reinforcement learning (RL) has the potential to enable effective decision-making over long timescales. Existing approaches, while promising, have yet to realize the benefits of large-scale training. In this work, we identify and solve several key challenges in scaling online hierarchical RL to high-throughput environments. We propose Scalable Option Learning (SOL), a highly scalable hierarchical policy gradient algorithm which achieves a ~35x higher throughput compared to existing hierarchical methods. To demonstrate SOL's performance and scalability, we train hierarchical agents using 30 billion frames of experience on the complex game of NetHack, significantly surpassing flat agents and demonstrating positive scaling trends. We also validate SOL on MiniHack and Mujoco environments, showcasing its general applicability.
General Machine Learning · Data
LLM-based agents already operate in production across many industries, yet we lack a clear understanding of which technical methods make these deployments successful. We present the first systematic study of Characterizing Agents in Production (CAP) using first-hand data from agent developers. We conducted 20 in-depth case studies through interviews and surveyed 306 practitioners across 26 domains. We examine why organizations build agents, how they build them, how they evaluate them, and the key challenges they face in deployment. Our findings show that production agents rely on simple, controllable approaches: 68% execute at most 10 steps before human intervention, 70% rely on prompting off-the-shelf models rather than weight tuning, and 74% depend primarily on human evaluation. Reliability—defined as consistent correct behavior over time—emerges as the dominant challenge, which practitioners address through system-level design choices. CAP documents the current state of production agents, providing the research community with visibility into real-world deployment practices and underexplored research opportunities.
Video language models (Video-LLMs) are prone to hallucinations, often generating plausible but ungrounded content when visual evidence is weak, ambiguous, or biased. Existing decoding methods, such as contrastive decoding (CD), rely on random perturbations to construct contrastive data for mitigating hallucination patterns. However, such a way is hard to control the visual cues that drive hallucination or well align with model weaknesses. We propose Model-aware Counterfactual Data based Contrastive Decoding (MACD), a new inference strategy that combines model-guided counterfactual construction with decoding. Our approach uses the Video-LLM’s own feedback to identify object regions most responsible for hallucination, generating targeted counterfactual inputs at the object level rather than arbitrary frame or temporal modifications. These model-aware counterfactual data is then integrated into CD to enforce evidence-grounded token selection during decoding. Experiments on EventHallusion, MVBench, Perception-test and Video-MME show that MACD consistently reduces hallucination while maintaining or improving task accuracy across diverse Video-LLMs, including Qwen and InternVL families. The method is especially effective in challenging scenarios involving small, occluded, or co-occurring objects. Our code and data will be publicly released.
Applications · Computer Vision
Unsupervised Domain Adaptation (UDA) person search aims to transfer a model trained on a labeled source domain to an unlabeled target domain without using target annotations. However, existing UDA methods frequently neglect the issue of scale inconsistency between the source and target domains. These inconsistency arises from different variations in Camera height, tilt angle, and focal length change. To address this challenge, we propose a Scale-Aware Consistent Alignment Learning (SCALE) framework. Specifically, we propose a Scale-aware Domain Harmonization (SDH) adaptively harmonizes semantic and structural scales through cross-path interaction and consistency refinement to alleviate cross-domain scale inconsistency. To further improve the pseudo-label inaccuracies, we introduce a Bidirectional Cluster Regularization (BCR) strategy, which obtains more reliable pseudo-labels by refining the results a second time. By collaboratively alleviating the impact of scale misalignment and enhancing pseudo-label reliability, our approach achieves state-of-the-art performance on two benchmark person search datasets, with 82.3% mAP and 84.0% top-1 on the CUHK-SYSU dataset, 41.7% mAP and 82.4% top-1 on the PRW dataset.
Applications · Neuroscience, Cognitive Science
Recent work has shown that scaling large language models (LLMs) improves their alignment with human brain activity, yet it remains unclear what drives these gains or which representational properties are responsible. Although larger models often yield better task performance and brain alignment, they are increasingly difficult to analyze mechanistically. This raises a fundamental question: \emph{what is the minimal model capacity required to capture brain-relevant representations?} To address this question, we systematically investigate how constraining model scale and numerical precision affects brain alignment. We compare full-precision LLMs, small language models (SLMs), and compressed variants (quantized and pruned) by predicting fMRI responses during naturalistic language comprehension. Across model families up to 14B parameters, we find that 3B SLMs achieve brain predictivity indistinguishable from larger LLMs, whereas 1B models degrade substantially, particularly in semantic language regions. Brain alignment is remarkably robust to compression: most quantization and pruning methods preserve neural predictivity, with GPTQ as a consistent exception. Linguistic probing reveals a dissociation between task performance and brain predictivity: compression degrades discourse, syntax, and morphology, yet brain predictivity remains largely unchanged. Overall, brain alignment saturates at modest model scales and is resilient to compression, challenging common assumptions about neural scaling and motivating compact models for brain-aligned language modeling.
General Machine Learning · Representation Learning
Latent Factorization-of-Tensors (LFT) models are an effective approach for relation extraction. Existing LFT models assume each mode of the target tensor corresponds to a entity set and the relationships between entity sets are bipartite graphs to explore the relationships among entities within a mode. However, when the topological structure of entities in a mode is known, for example, entities are ordered physical quantities, such as time or coordinates, the relation between such modes forms a more complicated system, i.e., aligned bipartite networks, and existing LFT models cannot accurately capture this structure. This work is the first to recognize and analyze this issue, and proposes an Asymmetric Latent Factorization-of-Tensors (ALFT) model to address it. ALFT can understand aligned bipartite networks in mode pairs of a tensor by imposing constraints between particular mode pairs in the tensor network. Experimental results on real-world datasets demonstrate the existence of this issue and confirm that the proposed ALFT model can effectively resolve it.
Applications · Chemistry, Physics, and Earth Sciences
Generating periodic data---such as fractional atomic coordinates in crystal structures and phase patterns in compressive light-field (CLF) displays---is challenging because wrap-around boundaries complicate probabilistic modeling and learning. While Bayesian Flow Networks (BFNs) offer a powerful generative framework with strictly additive accuracy in Euclidean space, existing periodic adaptations typically sacrifice additivity and become sensitive to schedule heuristics. We introduce \emph{PeriodicBFN}, which embeds each periodic scalar into a two-dimensional unit-circle representation and performs Gaussian Bayesian updates in the resulting Cartesian space, thereby restoring strictly additive accuracy. To address invariance in periodic generative modeling, we further derive a Rao--Blackwellized objective that analytically marginalizes global periodic translations, producing a translation-invariant target with reduced gradient variance. Experiments on crystal structure prediction and multi-layer phase synthesis for CLF displays demonstrate improved training stability and strong performance. To our knowledge, this is the first work to extend periodic-data generative modeling to phase synthesis for modern glasses-free 3D display systems.
Multimodal image fusion integrates complementary information from different modalities. However, large cross-modal discrepancies and local conflicts often introduce uncertainty into fusion decisions. This uncertainty can bias modality allocation in inconsistent regions, leading to information loss or the propagation of artifacts. Therefore, we address this problem by formally casting image fusion as an integrated probabilistic decision system that couples prior decision-making with posterior risk minimization. Based on this view, we propose a dIfference-aware Decision-lEArning muLtimodal image fusion paradigm (IDEAL). It treats cross-modal differences as decision triggers and learns contribution policies conditioned on local conditions. Specifically, we use a difference-attention module to generate multi-scale difference maps as spatial decision conditions. We also obtain spectral conditions by projecting features into the frequency domain, where power-spectrum energy, complementary spectra, and spectral-entropy reliability characterize modality discrepancy and reliability. We then employ a symmetric Beta prior to map these decision conditions to gating weights, yielding explicit and interpretable modality contribution policies. To improve robustness, we introduce an uncertainty modulation mechanism that reverts the policy to conservative mixing when conditions is insufficient. Extensive experiments demonstrate stable and competitive performance.
Deep Learning · Large Language Models
Multi-step theorem prediction is a central challenge in automated reasoning. Existing neural–symbolic approaches rely heavily on supervised parametric models, which exhibit limited generalization to evolving theorem libraries. In this work, we explore training-free theorem prediction through the lens of in-context learning (ICL). We identify a critical scalability bottleneck, termed Structural Drift: as reasoning depth increases, the performance of vanilla ICL degrades sharply, often collapsing to near zero. We attribute this failure to the LLM’s inability to recover latent topological dependencies, leading to unstructured exploration. To address this issue, we propose Theorem Precedence Graphs, which encode temporal dependencies from historical solution traces as directed graphs, and impose explicit topological constraints that effectively prune the search space during inference. Coupled with retrieval-augmented graph construction and a stepwise symbolic executor, our approach enables LLMs to act as structured planners without any gradient-based optimization. Experiments on the FormalGeo7k benchmark show that our method achieves 89.29\% accuracy, substantially outperforming ICL baselines and matching state-of-the-art supervised models. These results indicate that explicit structural priors offer a promising direction for scaling LLM-based symbolic reasoning.
Neural architecture design lacks first principles: innovations are discovered empirically and justified post-hoc, with no systematic way to diagnose *why* an architecture fails or derive *what* repair will succeed. We introduce the *Axiomatic Atlas*, encoding requirements as composable axioms over graph connectivity, operator contracts, numerical stability, and information preservation. Given an operator library and wiring conventions, the Atlas constructs certificates lower-bounding output variation via min-cut analysis and diagnoses failures by locating axiom violations. Crucially, the framework is prescriptive: each violation implies a targeted repair, reducing architecture design to constraint satisfaction. We prove variation bounds under exact and finite-precision arithmetic, enabling modular verification across transformers, MoEs, SSMs, and GNNs. Four Atlas-derived interventions validate the approach: +46 percentage points on GNN bottlenecks, $3\times$ robustness to MoE quantization, 83\% gap closure with adaptive expert budgets, and 0\%$\to$100\% retrieval via orthogonal keys---each against matched negative controls.
Deep Learning · Large Language Models
Preference-based reinforcement learning (RL) is a key paradigm for aligning policies with human judgments, yet its theoretical behavior in distributed settings where preference data are fragmented across heterogeneous users remains poorly understood. Direct Preference Optimization (DPO) avoids explicit reward modeling but lacks convergence guarantees under federated and decentralized training, where communication constraints and non-IID preferences fundamentally alter optimization dynamics. We provide the first convergence and time-complexity analysis of DPO in distributed environments. Modeling personalized offline RL with user-specific preference distributions, we characterize the induced global optimization landscape. For federated DPO, we derive convergence rates that quantify the impact of client drift, communication frequency, and preference heterogeneity; for decentralized DPO, we establish convergence over general communication graphs and show how spectral connectivity governs optimization speed and consensus. Our results lay a theoretical foundation for scalable and privacy-preserving distributed preference optimization.
Applications · Computer Vision
Texture recognition is often framed as matching an image to a static training-set dictionary or codebook. In practice, this assumption is brittle: label-preserving transformations (illumination, scale, compression, blur) can shift test features away from the fixed training dictionary, producing a training-set codebook misalignment that limits accuracy. We propose STD-Former (Simple Texture Dictionary Transformer), a lightweight framework for image-conditioned texture dictionary encoding. Instead of comparing against a static codebook, STD-Former extracts a compact set of Intrinsic Textons (dictionary atoms / codewords) from the input image itself, yielding self-aligned representations at inference. Our design is intentionally simple and uses a decoupled two-stage recipe. In Stage 1, a Texture Dictionary Extractor (TDE) is pre-trained with a self-supervised Texton Coverage Loss that encourages the learned textons to collectively cover the image patch feature manifold. In Stage 2, a classifier is trained on the encoded dictionary representation; optionally, we add a Sparse Topological Loss derived from 0D persistent homology, which is equivalent to supervising only the (B-1) edges of a minimum spanning tree (MST) in each batch, providing efficient structure regularization. Across six standard texture benchmarks, STD-Former and STD-Former+ achieve new state-of-the-art results.
Vision-Language-Action (VLA) models have recently shown strong decision-making capabilities in autonomous driving. However, existing VLAs often struggle with achieving efficient inference and generalizing to novel autonomous vehicle configurations and driving scenarios. In this paper, we propose Reasoning-VLA, a general and efficient action-generation VLA framework. The proposed model employs a set of learnable action queries, implicitly guided by predefined spatial representations to enhance spatial awareness. These learnable queries interact with reasoning-enhanced vision–language features to generate continuous action trajectories in parallel. To promote robust generalization, we consolidate eight publicly available autonomous driving datasets into a standardized, Chain-of-Thought reasoning–based, and easy-to-use data format for model training. Leveraging both supervised learning and reinforcement learning fine-tuning, extensive empirical evaluations across multiple benchmarks demonstrate that Reasoning-VLA achieves state-of-the-art performance, strong generalization capability, and the excellent inference speed with parallel decode.
Optimization · Large Scale, Parallel and Distributed
Federated learning~(FL) across multiple HPC facilities faces stochastic \emph{admission delays} from batch schedulers that dominate wall-clock time. Synchronous FL suffers from severe stragglers, while asynchronous FL accumulates stale updates when queues spike. We propose \fedqueue{}, a queue-aware FL protocol that incorporates scheduler delays directly into training and aggregation, which (i) predicts per-facility queue delays online to budget local work, (ii) applies cutoff-based admission that buffers late arrivals to bound staleness, and (iii) performs staleness-aware aggregation to stabilize heterogeneous local workloads. We prove the convergence for non-convex objectives at rate $\mathcal{O}(1/\sqrt{R})$ under bounded staleness, and show that the admission controls yield bounded staleness with high probability under queue-prediction error. Real-world cross-facility deployment of \fedqueue{} shows 20.5\% improvement over baseline algorithms. Controlled queue simulations demonstrate robust improvement over the baselines; in particular, about 34\% reduction in time to reach a target accuracy level under high queue variance and non-IID partitions.
Deep Learning · Algorithms
Deep learning optimization relies heavily on the assumption of smooth loss landscapes, a condition systematically violated by modern architectures due to non-smooth components like ReLU activations and quantization operators. In such non-smooth regimes, adaptive optimizers such as Adam suffer from gradient chattering—violent oscillations caused by conflicting signals within the Clarke subdifferential—leading to poor convergence and suboptimal generalization. To address this, we introduce Singularity-aware Adam (S-Adam), a novel optimizer that stabilizes training by dynamically modulating step sizes based on local geometric instability. Our key contribution is the Local Geometric Instability (LGI) metric, a computationally efficient estimator of the Clarke subdifferential diameter derived from the variance of randomized directional derivatives. S-Adam incorporates an adaptive damping mechanism $\exp(-\lambda \rho_t)$ that decelerates updates in high-instability regions while preserving fast convergence in smooth basins. We provide a rigorous convergence analysis using differential inclusions, proving that S-Adam converges almost surely to $(\delta, \epsilon)$-Clarke stationary points at the optimal $\mathcal{O}(1/\sqrt{T})$ rate. Empirical evaluations on Quantization-Aware Training (QAT) and high-volatility transfer learning demonstrate that S-Adam consistently outperforms AdamW and Prox-SGD, achieving accuracy gains of up to +6\% on CIFAR-100 and +3\% on TinyImageNet while effectively mitigating gradient oscillations.
Retrieval-augmented generation (RAG) systems commonly improve robustness via query-time adaptations such as query expansion and iterative retrieval. While effective, these approaches are inherently stateless: adaptations are recomputed for each query and discarded thereafter, precluding cumulative learning and repeatedly incurring inference-time cost. Index-side approaches like key expansion introduce persistence but rely on offline preprocessing or heuristic updates that are weakly aligned with downstream task utility, leading to semantic drift and noise accumulation. We propose Evolving Retrieval Memory (ERM), a training-free framework that transforms transient query-time gains into persistent retrieval improvements. ERM updates the retrieval index through correctness-gated feedback, selectively attributes atomic expansion signals to the document keys they benefit, and progressively evolves keys via stable, norm-bounded updates. We show that query and key expansion are theoretically equivalent under standard similarity functions and prove convergence of ERM’s selective updates, amortizing optimal query expansion into a stable index with zero inference-time overhead. Experiments on BEIR and BRIGHT across 13 domains demonstrate consistent gains in retrieval and generation, particularly on reasoning-intensive tasks, at native retrieval speed.
Theory · Reinforcement Learning and Planning
We present the first high-probability optimal regret bound for a policy optimization technique applied to the problem of stochastic contextual multi-armed bandit (CMAB) with general offline function approximation. Our algorithm is both efficient and achieves an optimal regret bound of $\widetilde{O}(\sqrt{ K|\mathcal{A}|\log|\mathcal{F}|})$, where $K$ is the number of rounds, $\mathcal{A}$ is the set of arms, and $\mathcal{F}$ is the function class used to approximate the losses. Our results bridge the gap between theory and practice, demonstrating that the widely used policy optimization methods for the contextual bandit problem can achieve a rigorously-proved optimal regret bound. We support our theoretical results with an empirical evaluation of our algorithm.
Deep Learning · Large Language Models
Test-time scaling improves LLM performance by generating multiple candidate solutions, yet token-level sampling requires temperature tuning that trades off diversity against stability. Fine-grained MoE, featuring hundreds of well-trained experts per layer and multi-expert activation per token, offers an unexplored alternative through its rich routing space. We empirically characterize fine-grained MoE routing and uncover an informative pattern: router scores exhibit a certain head of high-confidence experts followed by an uncertain tail of low-confidence candidates. While single-run greedy accuracy remains stable when fewer experts are activated, multi-sample pass@n degrades significantly—suggesting that the certain head governs core reasoning capability while the uncertain tail correlates with reasoning diversity. Motivated by these findings, we propose Expert-Sample, a training-free method that preserves high-confidence selections while injecting controlled stochasticity into the uncertain tail, enabling diverse generation without destabilizing outputs. Evaluated on multiple fine-grained MoE models across math, knowledge reasoning, and code tasks, Expert-Sample consistently improves pass@n and verification-based accuracy. On Qwen3-30B-A3B-Instruct evaluated on GPQA-Diamond with 32 parallel samples, pass@32 rises from 85.4% to 91.9%, and accuracy improves from 59.1% to 62.6% with Best-of-N verification.
Deep Learning · Attention Mechanisms
At the core of modern prompting techniques is contextual sensitivity, the ability of large language models to adapt their predictions based on inference-time context. Despite its central role, inference behavior under strong contextual influence remains poorly understood, particularly at the level of internal inference dynamics. To bridge this gap, we introduce a theoretical framework for analyzing contextual influence through inference dynamics, enabling quantitative characterization of inference behavior beyond output-level answer changes. Our analysis shows that inference dynamics do not exhibit unbounded drift under repeated contextual assertions. Instead, predictive representations converge to stable, query-dependent regimes that fundamentally constrain whether contextual signals can alter a model’s prediction. This leads to a surprising finding: Repeated contextual assertions do not act as accumulating evidence during inference and may therefore fail to alter a model’s prediction even under unbounded repetition, while in other cases a prediction change becomes inevitable. We empirically validate our theoretical predictions across diverse models and tasks, demonstrating strong alignment between theory and observed inference behavior. These contributions offer a principled pathway toward characterizing the limits of contextual influence during inference, and provide practical implications for designing and evaluating repetition-based prompting methods.