论文检索

Applications · Robotics

XIAOHAN YE, Kui Wu, Taku Komura, zherong pan

With the maturation of differentiable physics, its role in various downstream applications—such as model-predictive control, robotic design optimization, and neural PDE solvers—has become increasingly important. However, the derivative information provided by differentiable simulators can exhibit abrupt changes or vanish altogether, impeding the convergence of gradient-based optimizers. In this work, we demonstrate that such erratic gradient behavior is closely tied to the design of contact models. We further introduce a set of properties that a contact model must satisfy to ensure well-behaved gradient information. Lastly, we present a practical contact model for differentiable rigid-body simulators that satisfies all of these properties while maintaining computational efficiency. Our experiments show that, even from simple initializations, our contact model can discover complex, contact-rich control signals, enabling the successful execution of a range of downstream locomotion and manipulation tasks.

General Machine Learning · Representation Learning

Jihoon Tack, Jack Lanchantin, Jane Dwivedi-Yu, Andrew Cohen, Ilia Kulikov, Janice Lan, Shibo Hao, Yuandong Tian, Jason E Weston, Xian Li

Next token prediction has been the standard training objective used in large language model pretraining. Representations are learned as a result of optimizing for token-level perplexity. We propose Continuous Concept Mixing (CoCoMix), a novel pretraining framework that combines discrete next token prediction with continuous concepts. Specifically, CoCoMix predicts ``continuous concepts'' learned from a pretrained sparse autoencoder and mixes them into the model's hidden state by interleaving with token hidden representations. Through experiments on multiple benchmarks, including language modeling and downstream reasoning tasks, we show that CoCoMix is more sample efficient and consistently outperforms standard next token prediction and knowledge distillation. We find that combining both concept learning and interleaving in an end-to-end framework is critical to performance gains. Furthermore, CoCoMix enhances interpretability and steerability by allowing direct inspection and modification of the predicted concept, offering a transparent way to guide the model’s internal reasoning process.

Optimization · Everything Else

Rui Zhang, Zhichao Lu

The rise of Large Language Model-based Automated Algorithm Design (LLM-AAD) has transformed algorithm development by autonomously generating code implementations of expert-level algorithms. Unlike traditional expert-driven algorithm development, in the LLM-AAD paradigm, the algorithm's ideas are often implicitly embedded in the generated code. Therefore, assessing algorithmic similarity directly from code, distinguishing genuine algorithmic innovation from mere syntactic variation, becomes essential. While code similarity metrics exist, they fail to capture algorithmic similarity, as they focus on surface-level syntax or output equivalence rather than problem-solving behavior. We propose BehaveSim, a novel method to measure algorithmic similarity through the lens of problem-solving trajectories (PSTrajs)—sequences of intermediate solutions produced during execution. By quantifying the alignment between PSTrajs using dynamic time warping (DTW), BehaveSim distinguishes algorithms with divergent logic despite syntactic or output-level similarities. We demonstrate its utility in two key applications: (i) Enhancing LLM-AAD: Integrating BehaveSim into existing LLM-AAD frameworks (e.g., FunSearch, EoH) promotes behavioral diversity, significantly improving performance on three AAD tasks. (ii) Algorithm analysis: BehaveSim clusters generated algorithms by behavior, enabling systematic analysis of problem-solving strategies—a crucial tool for the growing ecosystem of AI-generated algorithms. Data and code of this work are open-sourced at https://github.com/RayZhhh/behavesim.

Deep Learning · Generative Models and Autoencoders

Yinghui He, Abhishek Panigrahi, Yong Lin, Sanjeev Arora

Language models often show little to no improvement (i.e., “saturation”) when trained via vanilla supervised fine-tuning (SFT) on data similar to what they saw in their training set (e.g., MATH). We introduce a new fine-tuning strategy, STAT, to train such a student model by using the metacognition ability of a stronger large language model (LLM) as the teacher. The teacher uses the task dataset to create a list of skills needed for the task, and then labels each data point with its required skills (Didolkar et al., 2024). By monitoring the student’s answers, the teacher creates a Missing-Skill-Profile for the student, tracking how often they failed to apply each skill in their responses. We use this idea to build a modified training set in one of two ways. In STAT-Sel, the teacher uses an existing set of training examples but adaptively reweights them according to the Missing-Skill-Profile. In STAT-Syn, the teacher synthesizes additional examples involving missing skills. Across extensive experiments on Llama and Qwen models, our methods yield improvements of up to 7.5% on MATH, whereas SFT provides only limited gains. Furthermore, STAT enhances performance on out-of-distribution benchmarks (e.g., AIME24/25, AMC23, etc.) by an average of 4.6%. Crucially, we find that STAT is complementary to RL via GRPO (Shao et al., 2024): after the model is improved using STAT to address skill gaps, GRPO continues to add further gains. We conclude that skill-targeted adaptive training should broadly improve current training pipelines.

Computer Vision · Everything Else

Jinsung Lee, Suha Kwak

This paper studies the problem of data-specific neural representations, aiming for compact, flexible, and modality-agnostic storage of individual visual data using neural networks. Our approach considers a visual datum as a set of discrete observations of an underlying continuous signal, thus requiring models capable of capturing the inherent structure of the signal. For this purpose, we investigate state-space models (SSMs), which are well-suited for modeling latent signal dynamics. We first explore the appealing properties of SSMs for data-specific neural representation and then present a novel framework that integrates SSMs into the representation pipeline. The proposed framework achieved compact representations and strong reconstruction performance across a range of visual data formats, suggesting the potential of SSMs for data-specific neural representations.

Social Aspects · Trustworthy Machine Learning

Wenbo Pan, Jie Xu, Qiguang Chen, Junhao Dong, Libo Qin, Xinfeng Li, Yu Haining, Xiaohua Jia

Large Language Models (LLMs) should refuse to answer questions beyond their knowledge. This capability, which we term knowledge-aware refusal, is crucial for factual reliability, while existing metrics fail to capture this ability. In this work, we propose the Refusal Index (RI), a novel and principled metric that measures how accurately LLMs refuse questions they do not know. We define RI as Spearman's rank correlation between refusal probability and error probability. RI is practically measurable with a lightweight two-pass evaluation method which only require observed refusal rates across two standard evaluation runs. Extensive experiments across 16 models and 5 datasets demonstrate that RI accurately quantifies a model's knowledge-aware refusal capability. Notably, RI remains stable across different refusal rates and provides consistent model rankings independent of a model's overall accuracy and refusal rates. These properties suggest RI captures a stable, intrinsic aspect of model knowledge calibration. More importantly, RI provides insight into an important but previously overlooked aspect of LLM factuality: while LLMs achieve high accuracy on factual tasks, their refusal behavior can be unreliable and fragile.

Social Aspects · Trustworthy Machine Learning

Litian Liu, Reza Pourreza, Sunny Panchal, Apratim Bhattacharyya, Yubing Jian, Yao Qin, Roland Memisevic

Large Language Models (LLMs) are prone to generating plausible yet incorrect responses, known as hallucinations. Effectively detecting hallucinations is therefore crucial for the safe deployment of LLMs. Recent research has linked hallucinations to model uncertainty, suggesting that hallucinations can be detected by measuring dispersion over answer distributions obtained from multiple samples drawn from a model. While drawing from the distribution over tokens defined by the model is a natural way to obtain samples, in this work, we argue that it is sub-optimal for the purpose of detecting hallucinations. We show that detection can be improved significantly by taking into account model uncertainty in the Bayesian sense. To this end, we propose a very simple, training-free approach based on perturbing an appropriate subset of model parameters, or equivalently hidden unit activations, during sampling. We demonstrate that our approach significantly improves inference-time hallucination detection over standard sampling across diverse datasets, model architectures, and uncertainty metrics.

Computer Vision · Image and Video Generation

Cheng Cheng, Lin Song, Di An, Yicheng Xiao, Xuchong Zhang, Hongbin Sun, Ying Shan

Autoregressive (AR) models have emerged as a powerful framework for image generation, yet they remain bound by a fundamental limitation: once a prediction is made, it cannot be revised. Each step marches forward in a strict left-to-right sequence, causing small errors to accumulate and compromise the final image. In this work, we reimagine this process with TensorAR, a decoder-only AR model that shifts from predicting discrete tokens to predicting overlapping tensor windows. This simple change transforms image synthesis into a process of next-tensor prediction, enabling the model to refine earlier outputs while preserving the causal structure that defines autoregression. To guard against information leakage during training, we introduce a discrete tensor noising mechanism inspired by discrete diffusion theory, which injects categorical noise into input tensors. TensorAR is designed to be plug-and-play: unlike masked AR methods, it requires no architectural modifications, and unlike autoregressive diffusion, it preserves the familiar AR training paradigm. We evaluate TensorAR across both class-to-image and text-to-image tasks, showing consistent gains in generation quality and instruction-following ability, while achieving a superior balance between quality and latency. In doing so, TensorAR offers a new path forward for autoregressive generation---one where predictions are not just produced, but continually refined.

Reinforcement Learning · Batch/Offline

Qiushui Xu, Yuhao Huang, Yushu Jiang, Wenliang Zheng, Lei Song, Jinyu Wang, Jiang Bian

Accurate estimation of the Q-function is a central challenge in offline reinforcement learning. However, existing approaches often rely on a shared global Q-function, which is inadequate for capturing the compositional structure of tasks that consist of diverse subtasks. We propose In-context Compositional Q-Learning (ICQL), an offline RL framework that formulates Q-learning as a contextual inference problem and uses linear Transformers to adaptively infer local Q-functions from retrieved transitions without explicit subtask labels. Theoretically, we show that, under two assumptions---linear approximability of the local Q-function and accurate inference of weights from retrieved context---ICQL achieves a bounded approximation error for the Q-function and enables near-optimal policy extraction. Empirically, ICQL substantially improves performance in offline settings, achieving gains of up to 16.4\% on kitchen tasks and up to 8.8\% and 6.3\% on MuJoCo and Adroit tasks, respectively. These results highlight the underexplored potential of in-context learning for robust and compositional value estimation and establish ICQL as a principled and effective framework for offline RL.

Applications · Robotics

Cunxin Fan, Xiaosong Jia, Yihang Sun, Yixiao Wang, Jianglan Wei, Ziyang Gong, Xiangyu Zhao, Masayoshi Tomizuka, Xue Yang, Junchi Yan 等

The rise of foundation models paves the way for generalist robot policies in the physical world. Existing methods relying on text-only instructions often struggle to generalize to unseen scenarios. We argue that interleaved image-text inputs offer richer and less biased context and enable robots to better handle unseen tasks with more versatile human-robot interaction. Building on this insight, we introduce Interleave-VLA, a robot learning paradigm extending interleaved image-text instructions from digital world to directly generating continuous action sequences in the physical world. Interleave-VLA offers a natural, flexible, and model-agnostic paradigm that extends state-of-the-art vision-language-action (VLA) models with minimal modifications while achieving strong zero-shot generalization. Interleave-VLA also includes an automatic pipeline that converts text instructions from Open X-Embodiment into interleaved image-text instructions, resulting in a large-scale real-world interleaved embodied dataset with 210k episodes. Comprehensive evaluation in simulation and the real world shows that Interleave-VLA offers two major benefits: (1) improves out-of-domain generalization to unseen objects by 2× compared to text input baselines, (2) supports flexible task interfaces and diverse instructions in a zero-shot manner, such as hand-drawn sketches. We attribute Interleave-VLA's strong zero-shot capability to the use of instruction images, which effectively mitigate hallucinations, and the inclusion of heterogeneous multimodal datasets, enriched with Internet-sourced images, offering potential for scalability. [Our project site](https://interleave-vla.github.io/Interleave-VLA-Anonymous/) has more information.

Reinforcement Learning · Deep RL

Guojian Zhan, Letian Tao, Pengcheng Wang, Yixiao Wang, Yuxin Chen, Yiheng Li, Hongyang Li, Masayoshi Tomizuka, Shengbo Li

Learning expressive and efficient policy functions is a promising direction in reinforcement learning (RL). While flow-based policies have recently proven effective in modeling complex action distributions with a fast deterministic sampling process, they still face a trade-off between expressiveness and computational burden, which is typically controlled by the number of flow steps. In this work, we propose mean velocity policy (MVP), a new generative policy function that models the mean velocity field to achieve the fastest one-step action generation. To ensure its high expressiveness, an instantaneous velocity constraint (IVC) is introduced on the mean velocity field during training. We theoretically prove that this design explicitly serves as a crucial boundary condition, thereby improving learning accuracy and enhancing policy expressiveness. Empirically, our MVP achieves state-of-the-art success rates across several challenging robotic manipulation tasks from Robomimic and OGBench. It also delivers substantial improvements in training and inference speed over existing flow-based policy baselines.

Applications · Robotics

Yixiao Wang, Mingxiao Huo, Zhixuan Liang, Yushi Du, Lingfeng Sun, Haotian Lin, Jinghuan Shang, Chensheng Peng, Mohit Bansal, Mingyu Ding 等

Pretrained vision foundation models (VFMs) advance robotic learning via rich visual representations, yet individual VFMs typically excel only in specific domains, limiting generality across tasks. Distilling multiple VFMs into a unified representation can mitigate this limitation but often yields inflexible task-specific feature selection and requires costly full retraining to incorporate robot-domain knowledge. We propose VER, a Vision Expert transformer for Robot learning. During pretraining, VER distills multiple VFMs into a vision expert library. We then fine-tune only a lightweight routing network (fewer than 0.4% of parameters) to dynamically select task-relevant experts from the pretrained library for downstream robot tasks. We further introduce Patchwise Expert Routing with Curriculum Top-K Annealing to improve both flexibility and precision of dynamic expert selection. Moreover, VER supports parameter-efficient finetuning for scalable expert utilization and robot-domain knowledge integration. Across 17 diverse robotic tasks and multiple policy heads, VER achieves state-of-the-art performance. We find that VER reduces large-norm outliers in task-irrelevant regions (e.g., background) and concentrates on task-critical regions. More visualizations and codes are available in https://yixiaowang7.github.io/ver_page/.

Applications · Chemistry and Drug Discovery

Jinjia Feng, Zhewei Wei, Taifeng Wang, Zongyang Qiu

Molecular representations that fully capture geometric parameters such as bond angles and torsion angles are crucial for accurately predicting important molecular properties including enzyme catalytic activity, drug bioactivity, and molecular spectral characteristics, as demonstrated by extensive studies. However, current molecular graph representation learning approaches represent molecular geometric parameters only indirectly through combinations of atoms and bonds, neglecting the spatial relationships and interactions between these higher-order geometric structures. In this paper, we propose \textbf{TetraGT} (\textbf{Tetra}hedral \textbf{G}eometry-Driven Explicit \textbf{T}oken Interactions with Graph Transformer), a novel architecture that directly models molecular geometric parameters. Based on the spatial solid geometry theory of face angle and dihedral angle inequality, TetraGT explicitly represents bond angles and torsion angles as structured tokens for the first time, directly reflecting their intrinsic role in determining the molecular conformational stability and properties. Through our designed spatial tetrahedral attention mechanism, TetraGT achieves highly selective direct communication between structural tokens. Experimental results demonstrate that TetraGT achieves superior performance on the PCQM4Mv2 and OC20 IS2RE benchmarks. We also apply our pre-trained TetraGT model to downstream tasks including QM9, PDBBind, Peptides and LIT-PCBA, demonstrating that TetraGT delivers excellent results in transfer learning scenarios and shows scalability with increasing molecular size.

Xiequn Wang, Zhan ZHUANG, Shengda Luo, Yu Zhang

We introduce MoSA, a new parameter-efficient fine-tuning (PEFT) method that replaces low-rank factorization with randomized, fine-grained sharing of weight updates. Each adapted weight matrix is constructed by broadcasting a small set of learned scalars over a fixed tessellation, a pre-defined group assignment of weight entries of the weight matrix, producing expressive changes under the same parameter budget as low-rank adaptation (LoRA). MoSA requires no architectural changes and can be merged into the base model for zero-overhead inference. Across diverse language understanding and generation tasks, MoSA matches or surpasses strong PEFT baselines under strictly matched budgets. Analyses and ablations indicate that non-local parameter sharing acts as an effective regularizer, and that grouping design and budget allocation govern the expressivity–efficiency trade-off. These results position MoSA as a simple, scalable alternative to LoRA. Our code is available at https://github.com/XiequnWang/MoSA-ICLR26.

Social Aspects · Trustworthy Machine Learning

Aashish Kolluri, Rishi Sharma, Manuel Costa, Boris Köpf, Tobias Nießen, Mark Russinovich, Shruti Tople, Santiago Zanella-Beguelin

Indirect prompt injection attacks threaten AI agents that execute consequential actions, motivating deterministic system-level defenses. Such defenses can provably block unsafe actions by enforcing confidentiality and integrity policies, but currently appear costly: they reduce task completion rates and increase token usage compared to probabilistic defenses. We argue that existing evaluations miss a key benefit of system-level defenses: reduced reliance on human oversight. We introduce autonomy metrics to quantify this benefit: the fraction of consequential actions an agent can execute without human-in-the-loop (HITL) approval while preserving security. To increase autonomy, we design a security-aware agent that (i) introduces richer HITL interactions, and (ii) explicitly plans for both task progress and policy compliance. We implement this agent design atop an existing information-flow control defense against prompt injection and evaluate it on the AgentDojo and WASP benchmarks. Experiments show that this approach yields higher autonomy without sacrificing utility (task completion).

Optimization · Sampling and Optimization

Arpan Mukherjee, Marcello Bullo, Debabrota Basu, Deniz Gunduz

While test-time scaling with verification has shown promise in improving the performance of large language models (LLMs), role of the verifier and its imperfections remain underexplored. The effect of verification manifests through interactions of three quantities: (i) the generator’s *coverage*, (ii) the verifier’s *region of convergence* (ROC), and (iii) the sampling algorithm’s *sub-optimality*. Though recent studies capture subsets of these factors, a unified framework quantifying the geometry of their interplay is missing. We frame verifiable test-time scaling as a transport problem. This characterizes the interaction of coverage, ROC, and sub-optimality, and uncovers that the sub-optimality-coverage curve exhibits three regimes. A *transport regime* — where sub-optimality increases with coverage, a *policy improvement regime* — where sub-optimality may decrease with coverage, depending on the verifier’s ROC, and a *saturation regime* — where sub-optimality plateaus, unaffected by coverage. We further propose and analyze two classes of sampling algorithms — *sequential* and *batched*, and examine how their computational complexities shape these trade-offs. Empirical results with `Qwen`, `Llama`, and `Gemma` models corroborate our theoretical findings.

Applications · Health

Gyubok Lee, Woosog Chay, Heeyoung Kwak, Yeong Kim, Haanju Yoo, Oksoon Jeong, Meong Hi Son, Edward Choi

Despite the impressive performance of LLM-powered agents, their adoption for Electronic Health Record (EHR) data access remains limited by the absence of benchmarks that adequately capture real-world clinical data access flows. In practice, two core challenges hinder deployment: query ambiguity from vague user questions and value mismatch between user terminology and database entries. To address this, we introduce EHR-ChatQA, an interactive database question answering benchmark that evaluates the end-to-end workflow of database agents: clarifying user questions, using tools to resolve value mismatches, and generating correct SQL to deliver accurate answers. To cover diverse patterns of query ambiguity and value mismatch, EHR-ChatQA assesses agents in a simulated environment with an LLM-based user across two interaction flows: Incremental Query Refinement (IncreQA), where users add constraints to existing queries, and Adaptive Query Refinement (AdaptQA), where users adjust their search goals mid-conversation. Experiments with state-of-the-art LLMs (e.g., o4-mini and Gemini-2.5-Flash) over five i.i.d. trials show that while the best-performing agents achieve Pass@5 of over 90% (at least one of five trials) on IncreQA and 60–70% on AdaptQA, their Pass^5 (consistent success across all five trials) is substantially lower, with gaps of up to about 60%. These results underscore the need to build agents that are not only performant but also robust for the safety-critical EHR domain. Finally, we provide diagnostic insights into common failure modes to guide future agent development. Our code and data are publicly available at https://github.com/glee4810/EHR-ChatQA.

Applications · Time Series

Dongik Park, Hyunwoo Ryu, Suahn Bae, Keondo Park, Hyung-Sin Kim

Imputing missing values in multivariate time series remains challenging, especially under diverse missing patterns and heavy missingness. Existing methods suffer from suboptimal performance as corrupted temporal features hinder effective cross-variable information transfer, amplifying reconstruction errors. Robust imputation requires both extracting temporal patterns from sparse observations within each variable and selectively transferring information across variables—yet current approaches excel at one while compromising the other. We introduce T1 (Time series imputation with 1-to-1 channel-head binding), a CNN-Transformer hybrid architecture that achieves robust imputation through Channel-Head Binding—a mechanism creating one-to-one correspondence between CNN channels and attention heads. This design enables selective information transfer: when missingness corrupts certain temporal patterns, their corresponding attention pathways adaptively down-weight based on remaining observable patterns while preserving reliable cross-variable connections through unaffected channels. Experiments on 11 benchmark datasets demonstrate that T1 achieves state-of-the-art performance, reducing MSE by 46% on average compared to the second-best baseline, with particularly strong gains under extreme sparsity (70% missing ratio). The model generalizes to unseen missing patterns without retraining and uses a consistent hyperparameter configuration across all datasets. The code is available at https://github.com/Oppenheimerdinger/T1.

Social Aspects · Accountability, Transparency and Interpretability

Ruizhe Li, CHEN CHEN, Yuchen Hu, Yanjun Gao, Xi Wang, Emine Yilmaz

Retrieval-Augmented Generation (RAG) leverages large language models (LLMs) combined with external contexts to enhance the accuracy and reliability of generated responses. However, reliably attributing generated content to specific context segments, context attribution, remains challenging due to the computationally intensive nature of current methods, which often require extensive fine-tuning or human annotation. In this work, we introduce a novel Jensen–Shannon Divergence driven method to Attribute Response to Context (ARC-JSD), enabling efficient and accurate identification of essential context sentences without additional fine-tuning, gradient-calculation or surrogate modelling. Evaluations on a wide range of RAG benchmarks, such as TyDi QA, Hotpot QA, and Musique, using instruction-tuned LLMs in different scales demonstrate superior accuracy and significant computational efficiency improvements compared to the previous baselines. Furthermore, our mechanistic analysis reveals specific attention heads and multilayer perceptron (MLP) layers responsible for context attribution, providing valuable insights into the internal workings of RAG models and how they affect RAG behaviours.

Optimization · Zero-order and Black-box Optimization

Yimeng Zeng, Natalie Maus, Haydn Jones, Jeffrey Tao, Fangping Wan, Marcelo Der Torossian Torres, Cesar de la Fuente, Ryan Marcus, Osbert Bastani, Jacob Gardner

In multi-task Bayesian optimization, the goal is to leverage experience from optimizing existing tasks to improve the efficiency of optimizing new ones. While approaches using multi-task Gaussian processes or deep kernel transfer exist, the performance improvement is marginal when scaling beyond a moderate number of tasks. We introduce **BOLT**, an initialization-only transfer strategy that distills prior BO runs into an LLM which proposes candidates for new tasks, while the surrogate at test time remains single-task. The LLM is periodically fine-tuned on top solutions from completed runs, creating a closed loop where better BO outputs yield better initializations over time. This decoupled design scales to roughly 1500 tasks without the saturation observed for shared-surrogate MTBO and adds only a small, amortized overhead relative to the BO inner loops. We evaluate on two domains: database query optimization and antimicrobial peptide design. We demonstrate that LLM-generated initializations steadily improve and accelerate BO, and with sufficient fine-tuning, a few LLM samples often match or surpass full ''from-scratch'' BO with far fewer oracle calls.