论文检索

Social Aspects · Safety

Liyan Chen, Zoe Xi, Yael Kalai

As AI models continue to develop powerful capabilities, it becomes critical that we are able to verify that their output is aligned with our intentions. A recent line of work focuses on verification via debate, a model of interactive proofs where two competing powerful provers, or AI models, debate each other to convince a weak verifier, or a human, of the correctness of their claim. However, debate assumes that the two AI models possess equal abilities and that one of them is truthful, which may not be realistic. In this work, we show *how to avoid debate*: we initiate the study of *single-prover* interactive proofs for AI safety. Prior results in single-prover interactive proofs do not immediately carry over to the AI safety setting because they do not work when the computation has access to an oracle, such as to human judgment or an external database such as the web. We present doubly-efficient single-prover interactive proofs for oracle-aided computations (also known as relativizing proofs), in the settings where (1) the computation is robust, in the sense that the output does not change if at most a small fraction of the answers to oracle queries are incorrect, or (2) the oracle is a low-degree polynomial. These results suggest that interactive verification is possible even without debate, under structured or noise-tolerant oracle access.

Deep Learning · Large Language Models

Peter Chen, Xiaopeng Li, Xi Chen, Tianyi Lin

Direct alignment methods are increasingly used to align large language models (LLMs) with human preferences. However, many real-world alignment problems involve multiple conflicting objectives, where naive aggregation of preferences can lead to unstable training and poor trade-offs. In particular, weighted loss methods may fail to identify update directions that simultaneously improve all objectives, and existing multi-objective approaches often rely on explicit reward models, introducing additional complexity and distorting user-specified preferences. The contributions of this paper are two-fold. First, we propose a **R**eward-free **A**lignment framework for **C**onflicted **O**bjectives (RACO) that directly leverages pairwise preference data and resolves gradient conflicts via a novel clipped variant of conflict-averse gradient descent. We provide convergence guarantees to Pareto-critical points that respect user-specified objective weights, and further show that clipping can strictly improve convergence rate in the two-objective setting. Second, we improve our method using some heuristics and conduct experiments to demonstrate the compatibility of the proposed framework for LLM alignment. Both qualitative and quantitative evaluations on multi-objective summarization and safety alignment tasks across multiple LLM families (Qwen 3, Llama 3, Gemma 3) show that our method consistently achieves better Pareto trade-offs compared to existing multi-objective alignment baselines.

Deep Learning · Theory

Catherine Chen, Jingyan Shen, Xinyu Yang, Lihua Lei

We present an online, distribution-free framework for controlling the Conditional Value-at-Risk ($\operatorname{CVaR}$), extending conformal tail risk control to non-stationary and adversarial environments. Unlike classical risk control methods, which rely on stationarity or linearity of expectation, our approach provides provable safety guarantees for a nonlinear tail risk functional under arbitrary data generating processes that may drift or shift strategically over time. By leveraging deep connections between conformal tail risk control, parameter-free online learning, and the variational representation of $\operatorname{CVaR}$ introduced by Rockafellar and Uryasev, we develop a novel procedure for online $\operatorname{CVaR}$ control with adversarial regret guarantees. The proposed method operates without assumptions on the underlying data-generating process, making it broadly applicable in modern high-stakes deployment settings. We prove that the realized empirical $\operatorname{CVaR}$ is always controlled at the target level, and that the resulting control is asymptotically tight up to a vanishing $\tilde{O}(1/\sqrt{T})$ conservatism gap. We demonstrate the effectiveness of our approach on portfolio risk management and toxicity mitigation for Large Language Models (LLMs), where rare but catastrophic failures dominate system risk.

Gang Cao, Junying Zhang

Text-to-image generation is widely used, but many applications require strict instance-level layout alignment. Masked Autoregressive (MAR) models on continuous latents are efficient and high-fidelity, yet flattening 2D latents into 1D sequences weakens spatial topology and hinders precise control. We propose Structure-Aware RoPE-MAR (StructMAR) to turn layout alignment from soft correlation into explicit structural alignment. StructMAR integrates 2D Rotary Positional Embeddings with a Layout-Guided Attention Bias to mechanistically enforce token-to-instance correspondence. We further apply Group Relative Policy Optimization (GRPO) to better align training objectives with layout-centric evaluation. On COCO-Position, StructMAR achieves state-of-the-art alignment (57.2 AP, 79.4 mIoU) while maintaining image quality comparable to strong diffusion baselines. On COCO-MIG, it improves robustness in dense settings (ISR 61.9, mIoU 57.4) and achieves a 4.05$\times$ inference speedup. These results highlight the importance of explicit structural inductive biases for robust, efficient controllable autoregressive generation; code is available at https://anonymous.4open.science/r/StructMAR-FE92/.

Social Aspects · Accountability, Transparency, and Interpretability

Christina Lu, Jack Gallagher, Jonathan Michala, Kyle Fish, Jack Lindsey

Large language models can represent a variety of personas but typically default to a helpful Assistant identity cultivated during post-training. Across several different models, we find an “Assistant Axis" in their activation space, which captures the extent to which a model is operating in its default Assistant mode. Steering towards the Assistant direction reinforces helpful and harmless behavior; steering away increases the model’s tendency to identify as other entities. Measuring deviations along the Assistant Axis predicts “persona drift,” a phenomenon where models slip into exhibiting harmful or bizarre behaviors that are uncharacteristic of their typical persona. We find that persona drift is often driven by conversations demanding meta-reflection on the model’s processes or featuring emotionally vulnerable users. We show that restricting activations to a fixed region along the Assistant Axis can stabilize model behavior in these scenarios—and also in the face of adversarial persona-based jailbreaks. Our results suggest that post-training steers models toward a particular region of persona space but only loosely tethers them to it, motivating work on training and steering strategies that more deeply anchor models to a coherent persona.

Reinforcement Learning · Batch/Offline

Hyunwoo Kim, Hyo Kyung Lee

Off-policy reinforcement learning suffers from extrapolation errors when a learned policy selects actions that are weakly supported in the replay buffer. In this study, we address this issue by drawing an analogy to static friction in classical mechanics. From this perspective, the replay buffer is represented as a smooth, low-dimensional action manifold, where the support directions correspond to the tangential component, while the normal component captures the dominant first-order extrapolation error. This decomposition reveals an intrinsic anisotropy in value sensitivity that naturally induces a stability condition analogous to a friction threshold. To mitigate deviations toward unsupported actions, we propose Frictional Q-Learning, an off-policy algorithm that encodes supported actions as tangent directions using a contrastive variational autoencoder. We further show that an orthonormal basis of the orthogonal complement corresponds to normal components under mild local isometry assumptions. Empirical results on standard continuous-control benchmarks demonstrate robust, stable performance compared with existing baselines.

Deep Learning · Generative Models and Autoencoders

Yexin Liu, Wenjie Shu, Zile Huang, Haoze Zheng, Yueze Wang, Manyuan Zhang, Jinjing Zhu, Ser-Nam Lim, Harry Yang

Text-guided image-to-video generation has made substantial progress, yet it still struggles to execute text-specified edits that require substantial changes to a reference image (e.g., object addition, deletion, or modification). Empirically, our analysis reveals that this stems from **visual dominance**, where the reference image causes severe attention dispersion, inhibiting the model's ability to incorporate new semantic information. To address this, we propose **AlignVid**, a training-free intervention that re-calibrates the model's internal attention distribution. Drawing on an energy-based perspective of attention, AlignVid employs Attention Scaling Modulation (**ASM**) to reduce attention entropy and concentrate focus on semantic tokens, alongside Guidance Scheduling (**GS**) to maintain generation stability. To rigorously assess this capability, we present **OmitI2V**, a comprehensive benchmark for evaluating prompt adherence across object addition, deletion, and modification. Extensive experiments demonstrate that AlignVid effectively enhances semantic fidelity with negligible computational overhead.

Reinforcement Learning · Batch/Offline

Yisen Zhao, Peixi Peng, Xinyu Hu, Cong Li, Zhan Su, Zhuojian Li

Offline meta-reinforcement learning requires agents to generalize to unseen tasks from fixed datasets, yet existing sequence-based and MoE-based methods rely on implicit or token-level routing signals that fail to capture task-level structure. We propose the **Task-Guided Router (TGR)**, a structured expert-routing framework that explicitly models inter-task relationships via multi-view task representations that combine semantic descriptors, behavioral summaries, and latent dynamics features. Using structure-guided routing, TGR assigns experts based on global task compatibility rather than local trajectory fragments, enabling stable specialization and effective knowledge transfer across tasks.Extensive experiments on continuous-control benchmarks demonstrate that TGR consistently outperforms state-of-the-art offline meta-RL methods in few-shot generalization, particularly under sparse data and heterogeneous dynamics. Our results highlight the importance of task-level priors for robust offline meta-reinforcement learning.

General Machine Learning · Causality

Henri Arno, Dennis Frauen, Emil Javurek, Thomas Demeester, Stefan Feuerriegel

Many decision-making problems require ranking individuals by their treatment effects rather than estimating the exact effect magnitudes. Examples include prioritizing patients for preventive care interventions, or ranking customers by the expected incremental impact of an advertisement. Surprisingly, while causal effect estimation has received substantial attention in the literature, the problem of directly learning *rankings of treatment effects* has largely remained unexplored. In this paper, we introduce *Rank-Learner*, a novel two-stage learner that directly learns the ranking of treatment effects from observational data. We first show that naive approaches based on precise treatment effect estimation solve a harder problem than necessary for ranking, while our *Rank-Learner* optimizes a pairwise learning objective that recovers the true treatment effect ordering, without explicit CATE estimation. We further show that our *Rank-Learner* is Neyman-orthogonal and thus comes with strong theoretical guarantees, including robustness to estimation errors in the nuisance functions. In addition, our *Rank-Learner* is model-agnostic, and can be instantiated with arbitrary machine learning models (e.g., neural networks). We demonstrate the effectiveness of our method through extensive experiments where *Rank-Learner* consistently outperforms standard CATE estimators and non-orthogonal ranking methods. Overall, we provide practitioners with a new, orthogonal two-stage learner for ranking individuals by their treatment effects.

Deep Learning · Large Language Models

Miso Choi, Seonga Choi, Mincheol Kwon, Woosung Joung, Jinkyu Kim, Jungbeom Lee

Recent advances in large language models (LLMs) have led to the emergence of specialized multimodal LLMs (MLLMs), forming distinct model families that share a common foundation language models. Despite this evolutionary trend, it remains unexplored whether a fundamental behavioral link exists between derived MLLMs and their foundational LLMs. This work investigates the inheritance of truthfulness traits along this trajectory by quantifying the degree of context-truthfulness across individual attention heads. Our analysis of the Vicuna and Qwen families reveals a striking finding: MLLMs maintain a high correlation in truthfulness scores with their base LLMs, even after multi-modal fine-tuning and when evaluated on disparate data sources. Building on this insight, we propose a Soft Gating strategy that utilizes these inherited Truth Scores to amplify the influence of context-truthful heads while preserving the contributions of other heads. We validate our approach on base LLMs on HaluEval benchmark to demonstrate improved truthful reasoning. Subsequently, we show that Truth Scores derived from a base LLM can be effectively transferred to its multimodal descendants as a plug-and-play gate, achieving performance gains on POPE and CHAIR benchmark comparable to probing the MLLMs directly. Our work highlights a novel, systemic approach to enhancing reliability across an entire model family by leveraging its inherent, inherited traits.

Applications · Everything Else

Haoyi Zhang, Kairong Guo, Bojie Zhang, Yibo Lin, Runsheng Wang

Standard cells form the building blocks of digital circuits, so their delay and power critically influence chip-level performance; yet characterization (can be understood as evaluation of cell delay and power) still relies on slow simulation sweeps, and many fast predictors ignore layout geometry, missing coupling and layout-dependent effects. The challenge is to jointly represent layout geometry and netlist topology so models capture fine-grained spatial details together with structural connectivity for accurate performance prediction. We introduce \textbf{FusionCell}, a dual-modality predictor that treats routed layout geometry and netlist topology as inputs and fuses them explicitly in a unified model. A DeiT encoder processes three-layer routed layouts, while a graph transformer models heterogeneous device/net graphs. The modalities are integrated through a \textbf{topology-guided} mechanism, where the netlist acts as a structural ``map'' to actively query relevant physical regions in the layout for joint geometric and topological reasoning. We build a 7nm dataset based on the ASAP7 PDK with over 19.5k cells spanning 149 types using automatic tools, targeting six metrics: signal rise/fall delay, transition, and power. Experimental results demonstrate that \textbf{FusionCell} reduces regression error (average MAPE 0.92\%) and improves Spearman/Kendall ranking over baselines, while accelerating the characterization process by orders of magnitude compared to circuit simulation.

General Machine Learning · Causality

Emre Kavak, Tom Nuno Wolf, Christian Wachinger

Dataset bias often leads deep learning models to exploit spurious correlations instead of task-relevant signals. We introduce the Standard Anti-Causal Model (SAM), a unifying causal framework that characterizes bias mechanisms and yields a conditional independence criterion for causal stability. Building on this theory, we propose DISCO$_m$ and sDISCO, efficient and scalable estimators of conditional distance correlation that enable independence regularization in black-box models. Across six diverse datasets, our methods consistently outperform or are competitive in existing bias mitigation approaches, while requiring fewer hyperparameters and scaling seamlessly to multi-bias scenarios. This work bridges causal theory and practical deep learning, providing both a principled foundation and effective tools for robust prediction.

General Machine Learning · Causality

Emre Kavak, Tom Nuno Wolf, Christian Wachinger

Dataset bias often leads deep learning models to exploit spurious correlations instead of task-relevant signals. We introduce the Standard Anti-Causal Model (SAM), a unifying causal framework that characterizes bias mechanisms and yields a conditional independence criterion for causal stability. Building on this theory, we propose DISCO$_m$ and sDISCO, efficient and scalable estimators of conditional distance correlation that enable independence regularization in black-box models. Across six diverse datasets, our methods consistently outperform or are competitive in existing bias mitigation approaches, while requiring fewer hyperparameters and scaling seamlessly to multi-bias scenarios. This work bridges causal theory and practical deep learning, providing both a principled foundation and effective tools for robust prediction.

Deep Learning · Graph Neural Networks

Félix Marcoccia, Cédric Adjih, Victor Fagoo, Paul Mühlethaler, Thomas Watteyne, Gilles de Saint Julien

We introduce NetDiff, a node-conditioned denoising diffusion model that generates directional link topologies and a two-slot transmit/receive parity for mobile ad hoc networks. Directional antennas can yield high throughput but require globally consistent link decisions under sector, interference, connectivity, and half-duplex constraints. NetDiff improves global coherence with Absolute Cross-Attentive Modulation (ACAM) tokens, which provide permutation-invariant global signals and help the model match graph-level counts (e.g., density and sector usage). We also propose partial diffusion to update an existing topology with a small number of denoising steps, enabling fast reconfiguration under mobility. NetDiff reaches over 95 \% of target performance with constant inference time, outperforms heuristic and omnidirectional baselines, and improves over a strong diffusion graph-transformer baseline on key metrics.

Applications · Time Series

Caiyi Yang, Chenglin Li, Hao Zhang, Weijia Lu, ZHIFEI YANG, Wenrui Dai, xiaodong Zhang, Xiaofeng Ma, Can Zhang, Junni Zou 等

Contrastive learning has advanced the representation learning across domains, yet its success relies on data augmentations that preserve semantic contents while providing the view diversities. Multivariate time series, however, are inherently noisy, non-stationary, and lack such intuitive semantic cues. Consequently, standard heuristic augmentations that ignore semantic parts may risk destroying critical temporal dependencies. Though some recent approaches attempt to isolate informative components, they typically rely on an implicit neural mechanism to infer semantics, thus limiting the interpretability and controllability. To address this, we propose ProSAR, an information-theoretic framework that leverages the explicit prototype alignment to guide semantic augmentations, and establish a feedback loop between the augmentation, contrastive learning, and prototype updates. Specifically, grounded in our proposed Prototype-Conditioned Information Bottleneck principle, we leverage the time-domain prototypes as explicit anchors to localize semantic segments, and develop a time–frequency augmentation strategy that retains prototype-consistent information while discarding noise. To promote semantically consistent prototypes for a reliable view generation, we design a dual-prototype loop where the augmented views are encoded into representations and then the learned representations are clustered to update latent prototypes, whose decoded feedback refines the time-domain prototypes for the next round of augmentation. Experiments on diverse time-series benchmarks demonstrate that ProSAR outperforms the other contrastive learning methods on downstream forecasting and classification tasks.

Applications · Computer Vision

Liupeng Li, Haoqian Kang, Zhenyu Lu, Jinpeng Wang, Bin Chen, Ke Chen, Yaowei Wang

High-resolution (HR) image perception presents a key bottleneck for multimodal large language models (MLLMs). While visual search offers a promising solution, existing methods struggle with the trade-off between coverage and efficiency. Visual expert-assisted search is efficient but prone to blind spots when proposals fail, whereas scan-based search guarantees coverage at the cost of computational redundancy and semantic fragmentation. To address this dilemma, we introduce CVSearch, a training-free adaptive framework that dynamically schedules search strategies via an Assess-then-Search workflow. Specifically, CVSearch first invokes expert-assisted search when global information is insufficient, and only triggers a novel semantic-aware scanning mechanism upon failure. Distinct from rigid grid partitioning, this efficient scanning paradigm incorporates Semantic Guided Adaptive Patching to decompose images into semantically consistent regions, effectively mitigating object fragmentation. Furthermore, we devise a Dynamic Bottom-Up Search strategy driven by a Visual Complexity prior to enable efficient and precise iterative exploration of local details. Extensive experiments on HR benchmarks demonstrate that CVSearch achieves state-of-the-art accuracy while substantially improving search efficiency. Faithful code and configurations will be released.

Zhenyu Lu, Liupeng Li, Jinpeng Wang, Haoqian Kang, Manyuan Zhang, Yan Feng, Ke Chen, Yaowei Wang

While Multimodal Large Language Models (MLLMs) have achieved remarkable progress in general visual understanding, they suffer from a fundamental geometric fragility: standard visual representations often degrade rapidly under changes in viewpoint and viewing distance. Our analysis identifies that existing paradigms, whether relying on input-level fusion or latent reconstruction, remain entangled with the view-dependent pixel grid, failing to decouple intrinsic 3D structure from extrinsic camera pose. To address this, we introduce AffIn-Space, a framework that enforces strict affine invariance to enable robust spatial understanding. Unlike implicit learning approaches, AffIn-Space introduces a two-stage explicit decoupling mechanism. First, it employs explicit geometric resampling by utilizing decomposed affine quantities (derived from pose features) to spatially align 3D features to a canonical state before fusion. Second, within the MLLM, we implement affine-invariant constraints via an orthogonal projection mechanism, which mathematically strips away pose-dependent noise from the hidden states while retaining recoverable geometric semantics through conditional reconstruction. Extensive experiments on VSI-Bench, ScanQA, SQA3D, Scan2Cap, and EmbodiedScan demonstrate that AffIn-Space achieves state-of-the-art performance. Code and detailed instructions will be publicly released. Crucially, our approach exhibits superior stability against affine perturbations, validating the effectiveness of explicitly modeling geometric invariance for complex spatial tasks. Code will be made available. Extensive experiments show that AffIn-Space achieves state-of-the-art performance on spatial reasoning tasks (VSI-Bench, SQA3D and Scan2Cap), and on spatial grounding tasks (ScanRefer and EmbodiedScan), demonstrating the effectiveness of affine invariant representations for complex spatial understanding.

Deep Learning · Algorithms

Shenghao Yang, Zhichao Wang, Oleg Balabanov, N. Benjamin Erichson, Michael Mahoney

Matrix functions such as square root, inverse roots, and orthogonalization play a central role in preconditioned gradient methods for neural network training. This has motivated the development of iterative algorithms that avoid explicit eigendecompositions and rely primarily on matrix multiplications, making them well suited for modern GPU accelerators. We present PRISM (Polynomial-fitting and Randomized Iterative Sketching for Matrix functions computation), a general framework for accelerating iterative algorithms for computing matrix functions. PRISM combines adaptive polynomial approximation with randomized sketching: at each iteration, it fits a polynomial surrogate to the current spectrum via a sketched least-squares problem, adapting to the instance at hand with minimal overhead. We apply PRISM to accelerate Newton–Schulz-like iterations for matrix square roots and orthogonalization, which are core primitives in machine learning. Unlike prior methods, PRISM requires no explicit spectral bounds or singular value estimates; it adapts automatically to the evolving spectrum. Empirically, PRISM accelerates training when integrated into Shampoo and Muon optimizers.

Deep Learning · Large Language Models

Yi Ding, Ziliang Qiu, Bolian Li, Ruqi Zhang

Self-correction is essential for solving complex reasoning problems in vision–language models (VLMs), yet existing reinforcement learning (RL) methods struggle to learn it. Effective self-correction behaviors emerge only rarely during RL, making learning signals sparse. To address this challenge, we propose c**o**rre**ct**i**o**n-s**p**ecific rollo**u**t**s**} (**Octopus**), a rollout-augmentation framework that synthesizes dense self-correction supervision by recombining existing rollouts without computational overhead. This rollout augmentation simultaneously improves sample efficiency and stabilizes RL optimization. Furthermore, we introduce a two-stage RL training strategy that disentangles self-correction and direct reasoning, avoiding signal conflicts and enabling both behaviors to be learned effectively. Building on this, we introduce $\texttt{Octopus-8B}$, an advanced reasoning VLM with controllable self-correction capabilities. It achieves SoTA performance among open-source VLMs across 7 benchmarks, outperforming the best RLVR baseline by 1.0 score while requiring only $0.72\times$ training time per step.

Manuel Glöckler, Jose Pedro JP Manzano-Patron, Stamatios Sotiropoulos, Cornelius Schröder, Jakob Macke

Simulation plays a central role in scientific discovery. In many applications, the bottleneck is no longer running a simulator—it is choosing among large families of plausible simulators, each corresponding to different forward models/hypotheses consistent with observations. Over large model families, classical Bayesian workflows for model selection are impractical. Furthermore, amortized model-selection methods typically hard-code a fixed model prior—or complexity penalty—at training time, requiring users to commit to a particular parsimony assumption before seeing the data. We introduce PRISM, a simulation-based encoder-decoder that infers a joint posterior over both discrete model structures and associated continuous parameters, while enabling test-time control of model complexity via a tunable model prior that the network is conditioned on. We show that PRISM scales to families with combinatorially many (up to billions) of model instantiations on a synthetic symbolic regression task. As a scientific application, we evaluate PRISM on biophysical modeling for diffusion MRI data, showing the ability to perform model selection across several multi-compartment models, on both synthetic and in-vivo neuroimaging data.