论文检索

Deep Learning · Everything Else

Ming Ma, Jue Zhang, Fangkai Yang, Yu Kang, Qingwei Lin, Saravan Rajmohan, Dongmei Zhang

Large language model (LLM)–based multi-agent systems are challenging to debug because failures often arise from long, branching interaction traces. The prevailing practice is to leverage LLMs for log-based failure localization, attributing errors to a specific agent and step. However, this paradigm has two key limitations: (i) log-only debugging lacks validation, producing untested hypotheses, and (ii) single-step or single-agent attribution is often ill-posed, as we find that multiple distinct interventions can independently repair the failed task. To address the first limitation, we introduce DoVer, an intervention-driven debugging framework, which augments hypothesis generation with active verification through targeted interventions (e.g., editing messages, altering plans). For the second limitation, rather than evaluating on attribution accuracy, we focus on measuring whether the system resolves the failure or makes quantifiable progress toward task success, reflecting a more outcome-oriented view of debugging. On the datasets derived from GAIA and AssistantBench, DoVer flips 18–28% of failed trials into successes, achieves up to 16% milestone progress, and validates or refutes 30-60% of failure hypotheses. Our findings highlight intervention as a practical mechanism for improving reliability in agentic systems and open opportunities for more robust, scalable debugging methods for LLM-based multi-agent systems.

Dongyang Liu, Gao Peng, David Liu, DU, Zhen Li, Qilong Wu, Xin Jin, Sihan Cao, Shifeng Zhang, Steven HOI 等

Diffusion model distillation has emerged as a powerful technique for creating efficient few-step and single-step generators. Among these, Distribution Matching Distillation (DMD) and its variants stand out for their impressive performance, which is widely attributed to their core mechanism of matching the student's output distribution to that of a pre-trained teacher model. In this work, we challenge this conventional understanding. Through a rigorous decomposition of the DMD training objective, we reveal that the primary driver of few-step generation is not the distribution matching term, but a previously overlooked component we identify as \textit{\textbf{C}FG \textbf{A}ugmentation} (\textbf{CA}). We demonstrate that this term acts as the core "engine" of distillation, while the \textbf{D}istribution \textbf{M}atching (\textbf{DM}) term functions as a "regularizer" that ensures training stability and mitigates artifacts. We further validate this decoupling by demonstrating that while the DM term is a highly effective regularizer, it is not unique; simpler non-parametric constraints or GAN-based objectives can serve the same stabilizing function, albeit with different trade-offs. This decoupling of labor between CA and DM also allows a more principled analysis of the properties of both terms, leading to a more systematic and in-depth understanding. This new understanding enables us to propose principled modifications to the distillation process, such as decoupling the noise schedules for the engine and the regularizer, leading to further performance gains.

Social Aspects · Trustworthy Machine Learning

Shiyu Ni, Keping Bi, Jiafeng Guo, Minghao Tang, Jingtong wu, Zengxin Han, Xueqi Cheng

Honesty alignment—the ability of large language models (LLMs) to recognize their knowledge boundaries and express calibrated confidence—is essential for trustworthy deployment. Existing methods either rely on training-free confidence estimation (e.g., token probabilities, self-consistency) or training-based calibration with correctness annotations. While effective, the latter demands costly, large-scale labeling. We introduce Elicitation-Then-Calibration (EliCal), a two-stage framework that first elicits internal confidence using inexpensive self-consistency supervision, then calibrates this confidence with a small set of correctness annotations. This design substantially reduces annotation requirements while improving generalization across tasks. To support a large-scale study, we release HonestyBench, a benchmark covering ten free-form QA datasets with 560k training and 70k evaluation instances annotated with correctness and self-consistency signals. Experiments show that EliCal achieves near-optimal alignment with only 1k correctness annotations ($\sim$0.18\% of full supervision) and better alignment performance on unseen MMLU tasks than the calibration-only baseline, offering a scalable solution toward universal honesty alignment in LLMs.

Applications · Language, Speech and Dialog

Tiancheng Li, Yuhang Wang, Zhiyang Chen, Zijun Wang, Liyuan Ma, Guo-Jun Qi

Prompt evolution algorithms offer a powerful paradigm for enhancing AI systems based on closed-source models, while few work explores whether aggregating results from multiple prompts to reach a consensus can further advance the system capability boundary. In this paper, we introduce Consensus-Evolve (C-Evolve), an evolutionary algorithm that discovers a group of prompts whose aggregated outputs after majority voting achieve optimal performance. More specifically, C-Evolve employs an island-based evolutionary algorithm to maintain population diversity, and prompts from distinct islands are selected to form groups to aggregate their outputs. The key difference from single individual evolution is a voting score, which evaluates each individual prompt's contribution within groups. We take this as the fitness score for evolution instead of individual performance. Consequently, C-Evolve is more likely to produce and maintain prompts with higher potential to form a high-performing group and eliminate low-performing ones, gradually improving the group performance after reaching consensus. Our method achieves state-of-the-art performance across a wide range of tasks, including both open-ended tasks like HotpotQA and closed-ended tasks like MATH. On Qwen3-8B, C-Evolve achieves 70.67\% on HotpotQA and 43.88\% on IFBench, which are 4.95\% and 2.73\% higher than GEPA, respectively. For GPT-4.1-mini, the accuracy on IFBench is further improved to 47.96\% and reaches 95.33\% in the MATH benchmark. These results demonstrate the C-Evolve's competitive performance.

Ruoxi Cheng, Hao-Xuan Ma, Weixin Wang, Ranjie Duan, Jiexi Liu, Xiaoshuang Jia, Simeng Qin, Xiaochun Cao, Yang Liu, Yang Liu

Alignment is vital for safely deploying large language models (LLMs). Existing techniques are either reward-based--train a reward model on preference pairs and optimize with reinforcement learning (RL)--or reward-free--directly fine-tune on ranked outputs. Recent research show that well-tuned reward-based pipelines remain the most robust, and single-response demonstrations can outperform pairwise preference data. However, there still exist two key challenges: (1) imbalanced safety dataset that overrepresent common hazards while neglecting long-tail threats; and (2) static reward models that ignore task difficulty, limiting optimization efficiency and attainable gains. To address these limitations, we propose DR-IRL, which Dynamically adjusts Rewards through Inverse Reinforcement Learning. We first train category‑specific reward models using a balanced safety dataset of seven harmful categories as demonstration via IRL. Then we enhance Group Relative Policy Optimization (GRPO) by introducing dynamic reward scaling--adjusting rewards by task difficulty--data-level hardness by text encoder cosine similarity, model-level responsiveness by reward gaps. Extensive experiments across various benchmarks and LLMs demonstrate that DR-IRL outperforms all baseline methods in safety alignment while maintaining usefulness.

Reinforcement Learning · Deep RL

Yonatan Ashlag, Mirco Mutti, Aviv Tamar, Kfir Y Levy

Sample efficiency is one of the main bottlenecks for optimal decision making via reinforcement learning. Pretraining a policy to maximize the entropy of the state visitation can substantially speedup reinforcement learning of downstream tasks. It is still an open question how to maximize the state entropy in POMDPs, where the true states of the environment, or their entropy, are not observed. In this work, we propose to maximize the entropy of a sufficient statistic of the history, which is called an information state. First, we show that a recursive latent model that predicts future observations is an information state in this setting. Then, we provide a practical algorithm, called LatEnt, to simultaneously learn the latent model and a latent-based policy maximizing the corresponding entropy objective from reward-free interactions with the POMDP. We empirically show that our approach induces higher state entropy than existing methods, which translates to better performance on downstream tasks. As a byproduct, we open-source PROBE, the first benchmark to test reward-free pretraining in POMDPs.

Badr Moufad, Yazid Janati el idrissi, Navid Bagheri Shouraki, Alain Oliviero Durmus, Thomas Hirtz, Eric Moulines, Jimmy Olsson

Diffusion models have emerged as powerful priors for image editing tasks such as inpainting and local modification, where the objective is to generate realistic content that remains consistent with observed regions. In particular, zero-shot approaches that leverage a pretrained diffusion model, without any retraining, have been shown to achieve highly effective reconstructions. However, state-of-the-art zero-shot methods typically rely on a sequence of surrogate likelihood functions, whose scores are used as proxies for the ideal score. This procedure however requires vector-Jacobian products through the denoiser at every reverse step, introducing significant memory and runtime overhead. To address this issue, we propose a new likelihood surrogate that yields simple and efficient to sample Gaussian posterior transitions, sidestepping the backpropagation through the denoiser network. Our extensive experiments show that our method achieves strong observation consistency compared with fine-tuned baselines and produces coherent, high-quality reconstructions, all while significantly reducing inference cost.

Deep Learning · Theory

Fan Chen, Audrey Huang, Noah Golowich, Sadhika Malladi, Adam Block, Jordan Ash, Akshay Krishnamurthy, Dylan Foster

Language models demonstrate remarkable abilities when pre-trained on large text corpora and fine-tuned for specific tasks, but how and why pre-training shapes the success of the final model remains poorly understood. Notably, although pre-training success is often quantified by cross entropy loss, cross entropy can be poorly predictive of downstream performance. Instead, we provide a theoretical perspective on this relationship through the lens of coverage, which quantifies the probability mass the pre-trained model places on high-quality responses and which is necessary and sufficient for post-training and test-time scaling methods like Best-of-N to succeed. Our main results develop an understanding of the coverage principle, a phenomenon whereby next-token prediction implicitly optimizes toward a model with good coverage. In particular, we uncover a mechanism that explains the power of coverage in predicting downstream performance: coverage generalizes faster than cross entropy, avoiding spurious dependence on problem dependent parameters such as the sequence length. We also study practical algorithmic interventions with provable benefits for improving coverage, including (i) model/checkpoint selection procedures, (ii) gradient normalization schemes, and (iii) test-time decoding strategies.

Deep Learning · Theory

Anthony Nguyen, Antonin Bertrand, Sylvie Le Hégarat-Mascle, Emanuel Aldea, Mohammed EL-KORSO, Renaud LUSTRAT, Franck FLORIN

Deep neural networks often remain overconfident on Out-of-Distribution (OoD) inputs. We revisit this problem through Riemannian information geometry. We model the network's predictions as a statistical manifold and find that OoD inputs exhibit higher local Fisher-Rao sensitivity. By quantifying this sensitivity with the trace of the Fisher Information Matrix (FIM), we derive a unifying geometric connection between two common OoD signals: feature magnitude and output uncertainty. Analyzing the limitations of this multiplicative form, we extend our analysis using a product manifold construction. This provides a theoretical framework for the robust additive scores used in state-of-the-art (SOTA) detectors and motivates our final, competitive method.

Deep Learning · Theory

Yuhao Liu, Yu Chen, Rui Hu, Longbo Huang

Stochastic interpolants offer a robust framework for continuously transforming samples between arbitrary data distributions via ordinary or stochastic differential equations (ODEs/SDEs), holding significant promise for generative modeling. While previous studies have analyzed the finite-time convergence rate of discrete-time implementations for SDEs, the ODE counterpart remains largely unexplored. In this work, we bridge this gap by presenting a rigorous finite-time convergence analysis of numerical implementations for ODEs in the framework of stochastic interpolants. We establish novel discrete-time total variation error bounds for two widely used numerical solvers: the first-order forward Euler method and the second-order Heun's method. Our analysis also yields optimized iteration complexity results and step size schedules that enhance computational efficiency. Notably, when specialized to the diffusion model setting, our theoretical guarantees for the second-order method improve upon prior results in terms of both smoothness requirements and dimensional dependence. Our theoretical findings are corroborated by numerical and image generation experiments, which validate the derived error bounds and complexity analyses.

Deep Learning · Theory

Noah Golowich, Allen Liu, Abhishek Shetty

A major problem in the study of large language models is to understand their inherent low-dimensional structure. We introduce an approach to study the low-dimensional structure of language models at a model-agnostic level: as sequential probabilistic models. We first empirically demonstrate that a wide range of modern language models exhibit low-rank structure: in particular, matrices built from the model's logits for varying sets of prompts and responses have low approximate rank. We then show that this low-rank structure can be leveraged for generation --- in particular, we can generate a response to a target prompt using a linear combination of the model's outputs on unrelated, or even nonsensical prompts. On the theoretical front, we observe that studying the approximate rank of language models in the sense discussed above yields a simple universal abstraction whose theoretical predictions parallel our experiments. We then analyze the representation power of the abstraction and give provable learning guarantees.

Xin Liu, Runsong Zhao, Pengcheng Huang, Xinyu Liu, Junyi Xiao, chunyang xiao, Tong Xiao, Shengxiang Gao, Zhengtao Yu, JingBo Zhu

Context compression is an advanced technique that accelerates large language model (LLM) inference by converting long inputs into compact representations. Existing methods primarily rely on autoencoding tasks to train special compression tokens to represent contextual semantics. While autoencoding tasks enable compression tokens to acquire compression capabilities, we remark that such capabilities potentially conflict with actual downstream task requirements, prevent the models from learning the features more beneficial for real-world usage. Based on this observation, we propose Semantic-Anchor Compression (SAC), a novel method that shifts from autoencoding task based compression to an architecture that is equipped with this compression capability \textit{a priori}. Instead of training models to compress contexts through autoencoding tasks, SAC directly selects so-called anchor tokens from the original context and aggregates contextual information into their key-value (KV) representations. To ensure that anchors can effectively collect information, SAC introduces two key designs: (1) anchor embedding, a learnable embedding vector attached to the selected anchor tokens to mark compression carriers and (2) bidirectional attention modification, which enables anchor tokens to integrate information from the entire context. Experimental results show that SAC consistently outperforms existing context compression methods across different compression ratios and model sizes on question-answering and long-context summarization tasks. Our data, model and code have been released at \href{https://github.com/lx-Meteors/SAC}{https://github.com/lx-Meteors/SAC}.

Deep Learning · Theory

Blake Bordelon, Mary Letey, Cengiz Pehlevan

We study in-context learning (ICL) of linear regression in a deep linear self-attention model, characterizing how performance depends on various computational and statistical resources (width, depth, number of training steps, batch size and data per context). In a joint limit where data dimension, context length, and residual stream width scale proportionally, we analyze the limiting asymptotics for three ICL settings: (1) isotropic covariates and tasks (ISO), (2) fixed and structured covariance (FS), and (3) where covariances are randomly rotated and structured (RRS). For ISO and FS settings, we find that depth only aids ICL performance if context length is limited. Alternatively, in the RRS setting where covariances change across contexts, increasing the depth leads to significant improvements in ICL, even at infinite context length. This provides a new solvable toy model of neural scaling laws which depends on both width and depth of a transformer and predicts an optimal transformer shapes as a function of compute.

Deep Learning · Generative Models and Autoencoders

Zekun Wang, Anant Gupta, Zihan Dong, Christopher MacLellan

Catastrophic forgetting remains a central obstacle for continual learning in neural models. Popular approaches---replay and elastic weight consolidation (EWC)---have limitations: replay requires a strong generator and is prone to distributional drift, while EWC implicitly assumes a shared optimum across tasks and typically uses a diagonal Fisher approximation. In this work, we study the gradient geometry of diffusion models, which can already produce high-quality replay data. We provide theoretical and empirical evidence that, in the low signal-to-noise ratio (SNR) regime, per-sample gradients become strongly collinear, yielding an empirical Fisher that is effectively rank-1 and aligned with the mean gradient. Leveraging this structure, we propose a rank-1 variant of EWC that is as cheap as the diagonal approximation yet captures the dominant curvature direction. We pair this penalty with a replay-based approach to encourage parameter sharing across tasks while mitigating drift. On class-incremental image generation datasets (MNIST, FashionMNIST, CIFAR-10, ImageNet-1k), our method consistently improves average FID and reduces forgetting relative to replay-only and diagonal-EWC baselines. In particular, forgetting is nearly eliminated on MNIST and FashionMNIST and is roughly halved on ImageNet-1k. These results suggest that diffusion models admit an approximately rank-1 Fisher. With a better Fisher estimate, EWC becomes a strong complement to replay: replay encourages parameter sharing across tasks, while EWC effectively constrains replay-induced drift.

Deep Learning · Everything Else

Pengfei He, Zhenwei Dai, Bing He, Hui Liu, Xianfeng Tang, Hanqing Lu, Juanhui Li, Jiayuan Ding, Subhabrata Mukherjee, Suhang Wang 等

Large language model (LLM)-based agents increasingly rely on tool use to complete real-world tasks. While existing works evaluate the LLMs' tool use capability, they largely focus on the final answers yet overlook the detailed tool usage trajectory, i.e., whether tools are selected, parameterized, and ordered correctly. We introduce TRAJECT-Bench, a trajectory-aware benchmark to comprehensively evaluate LLMs' tool use capability through diverse tasks with fine-grained evaluation metrics. TRAJECT-Bench pairs high-fidelity, executable tools across practical domains with tasks grounded in production-style APIs, and synthesizes trajectories that vary in breadth (parallel calls) and depth (interdependent chains). Besides final accuracy, TRAJECT-Bench also reports trajectory-level diagnostics, including tool selection and argument correctness, and dependency/order satisfaction. Analyses reveal failure modes such as similar tool confusion and parameter-blind selection, and scaling behavior with tool diversity and trajectory length where the bottleneck of transiting from short to mid-length trajectories is revealed, offering actionable guidance for LLMs' tool use.

Deep Learning · Robustness

pragatheeswaran vipulanandan, Kamal Premaratne, Dilip Sarkar

Large language models (LLMs) exhibit strong generative capabilities but remain vulnerable to confabulations, fluent yet unreliable outputs that vary arbitrarily even under identical prompts. Leveraging a quantum tensor network–based pipeline, we propose a quantum physics-inspired uncertainty quantification framework that accounts for the aleatoric uncertainty in token sequence probability for semantic equivalence-based clustering of LLM generations. In turn, this offers a principled and interpretable scheme for hallucination detection. We further introduce an entropy-maximization strategy that prioritizes high-certainty, semantically coherent outputs and highlights entropy regions where LLM decisions are likely to be unreliable, offering practical guidelines for when human oversight is warranted. We evaluate the robustness of our scheme under different generation lengths and quantization levels, dimensions overlooked in prior studies, demonstrating that our approach remains reliable even in resource-constrained deployments. A total of 116 experiments on TriviaQA, NQ, SVAMP, and SQuAD across multiple architectures (Mistral-7B, Mistral-7B-instruct, Falcon-rw-1b, LLaMA-3.2-1b, LLaMA-2-13b-chat, LLaMA-2-7b-chat, LLaMA-2-13b and LLaMA-2-7b) show consistent improvements in AUROC and AURAC over state-of-the-art baselines.

Computer Vision · Image and Video Generation

Xianglu Wang, Wanlin Zhang, Hu Ding

Diffusion models have recently emerged as a powerful class of generative models and have achieved state-of-the-art performance in various image synthesis tasks. However, training diffusion models generally requires large amounts of data and suffer from overfitting when the dataset size is limited. To address these limitations, we propose a novel method called WILD-Diffusion, which is inspired by Wasserstein Distributionally Robust Optimization (WDRO), an important and elegant mathematical formulation from robust optimization area. Specifically, WILD-Diffusion utilizes WDRO to iteratively generate new training samples within a Wasserstein distance based uncertainty set centered at the limited data data distribution. This carefully designed method can progressively augment the training set throughout the training process and effectively overcome the obstacles caused by the limited data issue. Moreover, we establish the convergence guarantee for our algorithm even though the mixture of diffusion process and WDRO brings significant challenges to our analysis in theory. Finally, we conduct a set of experiments to verify the effectiveness of our proposed method. With WILD-Diffusion, we can achieve more than a $10$% reduction in FID using only $20$% of the training data across different datasets. Moreover, our method can attain state-of-the-art FID with as few as $100$ images, both in pretrained and non-pretrained settings.

Andres Potapczynski, Alex Ali, Andrew Gordon Wilson

Matrix operations, such as linear solves, eigendecompositions, and log determinants, are foundational building blocks for any number of downstream applications. Therefore, any broadly capable learning system should be able to effectively approximate these operations in its internal representation. Accordingly, there is great motivation to study transformers for linear algebra --- for if transformers cannot even semi-competently perform matrix operations, then we cannot expect them to form a basis for a generally intelligent system. We demonstrate that current techniques developing transformers for linear algebra have striking failure modes, prohibitive scaling, and particularly poor out-of-distribution generalization to other matrix distributions, and matrices of different sizes. Investigating further, we find that current transformer approaches operate as statistical interpolators, rather than discovering algorithms that will generalize to matrices from other distributions. Based on our understanding of these limitations, we develop a sequence of interventions that substantially improve scaling and performance, including matrix embeddings through a learnable projection, linear attention, looping, and a data pre-training distribution of structured matrices. We term the resulting method the \emph{RangeFormer}, which we show has significantly improved scaling and performance on challenging OOD matrices from the \emph{matrix market}. Moreover, with RangeFormer we show for the first time that transformers can be successfully applied to downstream tasks that involve iterative matrix operations, including Gaussian process learning, and improving the sampling distribution of randomized methods.

Zhuowei Chen, Bingchuan Li, Tianxiang Ma, Lijie Liu, Mingcong Liu, Yunsheng Jiang, Li, Xinghui Li, Liyang Chen, SiYu Zhou 等

Subject-to-video generation has witnessed substantial progress in recent years. However, existing models still face significant challenges in faithfully following textual instructions. This limitation, commonly known as the copy-paste problem, arises from the widely used in-pair training paradigm. This approach inherently entangles subject identity with background and contextual attributes by sampling reference images from the same scene as the target video. To address this issue, we introduce \textbf{Phantom-Data, the first general-purpose cross-pair subject-to-video consistency dataset}, containing approximately one million identity-consistent pairs across diverse categories. Our dataset is constructed via a three-stage pipeline: (1) a general and input-aligned subject detection module, (2) large-scale cross-context subject retrieval from more than 53 million videos and 3 billion images, and (3) prior-guided identity verification to ensure visual consistency under contextual variation. Comprehensive experiments show that training with Phantom-Data significantly improves prompt alignment and visual quality while preserving identity consistency on par with in-pair baselines.

Deep Learning · Graph Neural Networks

Marco Pacini, Mircea Petrache, Bruno Lepri, Shubhendu Trivedi, Robin Walters

Universality results for equivariant neural networks remain rare. Those that do exist typically hold only in restrictive settings: either they rely on regular or higher-order tensor representations, leading to impractically high-dimensional hidden spaces, or they target specialized architectures, often confined to the invariant setting. This work develops a more general account. For invariant networks, we establish a universality theorem under separation constraints, showing that the addition of a fully connected readout layer secures approximation within the class of separation-constrained continuous functions. For equivariant networks, where results are even scarcer, we demonstrate that standard separability notions are inadequate and introduce the sharper criterion of *entry-wise separability*. We show that with sufficient depth or with the addition of appropriate readout layers, equivariant networks attain universality within the entry-wise separable regime. Together with prior results showing the failure of universality for shallow models, our findings identify depth and readout layers as a decisive mechanism for universality, additionally offering a unified perspective that subsumes and extends earlier specialized results.