ICML 2026 Papers

As real-world knowledge continues to evolve, the parametric knowledge acquired by multimodal models during pretraining becomes increasingly difficult to remain consistent with real-world knowledge. Existing research on multimodal knowledge updating focuses only on learning previously unknown knowledge, while overlooking the need to update knowledge that the model has already mastered but that later changes; moreover, evaluation is limited to the same modality, lacking a systematic analysis of cross-modal consistency. To address these issues, this paper proposes MMKU-Bench, a comprehensive evaluation benchmark for multimodal knowledge updating, which contains over 25k knowledge instances and more than 49k images, covering two scenarios, updated knowledge and unknown knowledge, thereby enabling comparative analysis of learning across different knowledge types. On this benchmark, we evaluate a variety of representative approaches, including supervised fine-tuning (SFT), reinforcement learning from human feedback (RLHF), and knowledge editing (KE). Experimental results show that SFT and RLHF are prone to catastrophic forgetting, while KE better preserve general capabilities but exhibit clear limitations in continual updating. Overall, MMKU-Bench provides a reliable and comprehensive evaluation benchmark for multimodal knowledge updating, advancing progress in this field. The code and dataset are available at https://github.com/baochenfu/MMKU-Bench.

Recent advances in video diffusion models have demonstrated remarkable generative capability, yet adapting these large pretrained text-to-video (T2V) models to video super‑resolution (VSR) typically encounters challenges, such as artifacts introduced by complex degradations in real-world scenarios and compromised fidelity due to the strong generative capacity of the powerful T2V models. We present WEVSR, a novel approach that adapts a pretrained flow-matching video diffusion transformer to VSR. First, we design a task-oriented adaptation strategy that leverages timestep sampling and noise augmentation to enhance detail restoration while preserving structural stability. Second, we propose a lightweight multi-level discrete wavelet transform (DWT) front-end for the VAE encoder, injecting explicit frequency priors into the latent space without modifying the pretrained decoder. Extensive experiments across multiple VSR benchmarks demonstrate that WEVSR achieves state-of-the-art performance against existing approaches. Code and models will be released here.

Neural population geometry shapes downstream computation. Recent empirical findings in neurobiology suggest that a hyperbolic structure underlies population activity in the hippocampus. Here we provide a theoretical framework for this phenomenon. First, we propose a plausible construction of hippocampal tuning curves that statistically induces hyperbolic geometry. Next, we establish a connection between neural decoding and associative memory by demonstrating that the Modern Hopfield Network update rule computes the minimum mean-squared-error (MMSE) estimator. Finally, we introduce a novel associative memory model defined in hyperbolic space that yields significantly larger capacity than leading models. Our results suggest that animals encode spatial information as a latent hyperbolic cognitive map, improving both memory capacity and decoding accuracy.

Recent advances in visual representation learning have seen the rise of clustering-based vision backbones, which adopt clustering as a core paradigm for feature extraction. However, existing clustering-based backbones typically rely on a single clustering algorithm, whose inherent inductive bias limits their representational capacity. To address this, we propose EnFormer, which embeds ensemble clustering as a core component of feature extraction. EnFormer structures feature extraction around two steps: (i) Ensemble Generation, where several differentiable base clustering methods are introduced to capture diverse semantic structures; and (ii) Consensus Aggregation, which employs a differentiable mechanism to fuse the results of all base clusterings to reconstruct refined visual features. Extensive experiments show that EnFormer consistently outperforms existing clustering-based backbones across core vision tasks, with higher performance and significantly improved throughput.

Code Large Language Models (CodeLLMs) have been widely adopted for Natural Language to Programming Language code generation, powering applications with large user bases. Their performance, however, varies sharply across programming languages (PLs) and is particularly suboptimal for low-resource PLs due to data scarcity, limiting their overall usability. In this work, we introduce CodeChemist, a simple yet effective, training-free test-time scaling framework that transfers the model's functional knowledge from high-resource to low-resource PLs via synthesized test cases, without relying on external models. Specifically, CodeChemist first applies multi-temperature hedged sampling to generate a pool of candidate solutions in the low-resource PL and synthesizes a set of test inputs. It then estimates uncertainty: when uncertainty is low, it selects the output via in-language majority voting; otherwise, it constructs cross-lingual I/O test oracles by executing high-resource reference programs and selects the candidate with the highest pass rate. Extensive experiments demonstrate that CodeChemist significantly outperforms existing test-time scaling methods, improving code generation for both low-resource PLs (e.g., Lua) and complex-syntax PLs (e.g., C++, Java) without retraining.

Model merging combines multiple fine-tuned models into a single model by $\textit{adding}$ their weight updates, providing a lightweight alternative to retraining. Existing methods primarily target resolving conflicts between task updates, leaving the failure mode of over-counting shared knowledge unaddressed. We show that when tasks share aligned spectral directions (i.e., overlapping singular vectors), a simple linear combination repeatedly accumulates these directions, inflating the singular values and biasing the merged model toward shared subspaces. To mitigate this issue, we propose Singular Value Calibration (SVC), a training-free and data-free post-processing method that quantifies subspace overlap and rescales inflated singular values to restore a balanced spectrum. Across vision and language benchmarks, SVC consistently improves strong merging baselines and achieves state-of-the-art performance. Furthermore, by modifying only the singular values, SVC improves the performance of Task Arithmetic by 13.0\%. Code is available at https://github.com/lyymuwu/SVC.

Large Vision-Language Models (LVLMs) have shown remarkable performance on a wide range of vision-language tasks. Despite this progress, they are still prone to hallucination, generating responses that are inconsistent with visual content. In this work, we find that LVLMs tend to hallucinate when they pay insufficient attention to the correct visual evidence and gradually forget it during the generation process. We empirically find that although LVLMs overall attend insufficiently to visual evidence, they exhibit sensitivity to the correct visual evidence in specific layers, with notable inter-layer discrepancy. Motivated by this observation, we propose a novel hallucination mitigation method that enhances visual evidence based on **I**nter-**L**ayer **V**isual **A**ttention **D**iscrepancy (**ILVAD**). Specifically, we obtain the attention weights from early generated tokens to visual tokens across layers and identify the tokens that are repeatedly activated as visual evidence, forming a saliency map. We then enhance attention to visual evidence during generation through the saliency map to reduce visual forgetting. In addition, we leverage the saliency map to obtain attention scores of generated text to visual evidence, in order to select and emphasize text tokens that are strongly grounded in visual evidence. Our method is training-free and plug-and-play. Multiple benchmark evaluations conducted on five recently released models show that our method can consistently mitigate hallucinations in different LVLMs over various architectures. Code is available at https://github.com/ytx-ML/ILVAD.

Modern diffusion/flow-based models for image generation typically exhibit two core characteristics: (i) using multi-step sampling, and (ii) operating in a latent space. Recent advances have made encouraging progress on each aspect individually, paving the way toward one-step diffusion/flow without latents. In this work, we take a further step towards this goal and propose "pixel MeanFlow" (pMF). Our core guideline is to formulate the network output space and the loss space separately. The network target is designed to be on a presumed low-dimensional image manifold (i.e., x-prediction), while the loss is defined via MeanFlow in the velocity space. We introduce a simple transformation between the image manifold and the average velocity field. In experiments, pMF achieves strong results for one-step latent-free generation on ImageNet at 256$\times$256 resolution (2.22 FID) and 512$\times$512 resolution (2.48 FID), filling a key missing piece in this regime. We hope that our study will further advance the boundaries of diffusion/flow-based generative models.

Most existing language model agentic systems today are built and optimized for large language models (e.g., GPT, Claude, Gemini) via API calls; while powerful, this approach faces several limitations including high token costs and privacy concerns for sensitive applications. We introduce **EffGen**, an open-source agentic framework optimized for small language models (SLMs) that enables effective, efficient, and secure local deployment. EffGen makes four major contributions: **(1) Enhanced tool-calling** with prompt optimization that compresses input prompts by up to 70-80% (and 57% on average across our benchmarks) while preserving task semantics, **(2) Intelligent task decomposition** that breaks complex queries into parallel or sequential subtasks based on dependencies, **(3) Complexity-based routing** using five factors to make smart pre-execution decisions, and **(4) Unified memory system** combining short-term, long-term, and vector-based storage. Additionally, EffGen unifies multiple agent protocols (MCP, A2A, ACP) for cross-protocol communication. Results on 13 benchmarks show EffGen outperforms LangChain, AutoGen, and Smolagents with **higher success rates**, **faster execution**, and **lower memory**. Our results reveal that **prompt optimization and complexity routing have complementary scaling behavior**: optimization benefits SLMs more (11.2% gain at 1.5B vs 2.4% at 32B), while routing benefits large models more (3.6% at 1.5B vs 7.9% at 32B), providing consistent gains across all scales when combined. EffGen is open-source under the Apache 2.0 License, with the code available at https://github.com/ctrl-gaurav/effGen, the Python package at https://pypi.org/project/effgen/ (`pip install effgen`), and the project website and documentation at https://effgen.org/ and https://docs.effgen.org/.

Locating files and functions requiring modification in large software repositories is challenging due to their scale and structural complexity. Existing LLM-based methods typically treat this as a repository-level retrieval task and rely on multiple auxiliary tools, which often overlook code execution logic and complicate model control. We propose RepoNavigator, an LLM agent equipped with a single execution-aware tool: jumping to the definition of an invoked symbol. This unified design reflects the actual flow of code execution while simplifying tool manipulation. RepoNavigator is trained end-to-end via Reinforcement Learning (RL) directly from a base pretrained model, without relying on closed-source distillation. Experiments demonstrate that RL-trained RepoNavigator achieves state-of-the-art performance, with the 7B model outperforming 14B baselines, the 14B model surpassing 32B competitors, and the 32B model exceeding closed-source models such as GPT-5 on most metrics. These results confirm that integrating a single, structurally grounded tool with RL training provides an efficient and scalable solution for repository-level issue localization.

Pre-trained diffusion models have emerged as powerful generative priors for both unconditional and conditional sample generation, yet their outputs often deviate from the characteristics of user-specific target data. Such mismatches are especially problematic in domain adaptation tasks, where only a few reference examples are available and retraining the diffusion model is infeasible. Existing inference-time guidance methods can adjust sampling trajectories, but they typically optimize surrogate objectives such as classifier likelihoods rather than directly aligning with the target distribution. We propose *MMD Guidance*, a training-free mechanism that augments the reverse diffusion process with gradients of the *Maximum Mean Discrepancy (MMD)* between generated samples and a reference dataset. MMD provides reliable distributional estimates from limited data, exhibits low variance in practice, and is efficiently differentiable, which makes it particularly well-suited for the guidance task. Our framework naturally extends to prompt-aware adaptation in conditional generation models via product kernels. Also, it can be applied with computational efficiency in latent diffusion models (LDMs), since guidance is applied in the latent space of the LDM. Experiments on synthetic and real-world benchmarks demonstrate that MMD Guidance can achieve distributional alignment while preserving sample fidelity. The project code is available at [https://github.com/matinamehdizadeh/MMD-Guidance](https://github.com/matinamehdizadeh/MMD-Guidance).

Lens flare removal is challenging due to the large spatial extent of flare artifacts and their entangle-ment with scene structures, while existing meth-ods heavily rely on large-scale paired data. We propose a semi-supervised flare removal frame-work that enables stable learning from unlabeled images by jointly addressing pseudo-label relia-bility and representation discrimination. We pro-pose an adaptive pseudo-label repository that pro-gressively refines pseudo supervision through no-reference quality assessment, momentum-based updates, and invalid label filtering, effectively mit-igating error accumulation. Moreover, we pro-pose a flare-aware contrastive loss that explic-itly treats flare-contaminated inputs as negatives and performs patch-level contrastive learning, en-couraging representations that are discriminative against flare patterns while remaining consistent with reliable pseudo targets. Extensive experi-ments on multiple flare benchmarks demonstrate that the proposed framework is model-agnostic and consistently improves performance and ro-bustness.

Multimodal large language models (MLLMs) are playing an increasingly important role across multiple domains. In many applications, such as robotics, it is crucial to quantify the uncertainty in the output of these models. } We develop Fused Uncertainty with Semantic Evidence (FUSE), a probabilistic framework for capturing two complementary sources of uncertainty in multimodal language modeling: (i) aleatoric embedding-level uncertainty derived from input data vision-language ambiguity, and (ii) epistemic model-level uncertainty estimated from the semantic response diversity of MLLMs. Our approach formulates a Bayesian fusion mechanism that analytically combines these uncertainty sources to produce a scalar measure of uncertainty. This measure serves as a novel uncertainty representation for downstream applications of MLLMs and provides a principled foundation for uncertainty calibration in multimodal systems, improving reliability and downstream performance in MLLM-based reasoning and vision-language tasks. We demonstrate that our method outperforms baselines in providing uncertainty estimates and achieves state-of-the-art uncertainty calibration.

An artificial intelligence must have a model of its environment that is *causal*, supporting reasoning about interventions and counterfactuals, and also *combinatorial*, supporting generalization to unseen combinations of objects. In this work, we formally study when and how such a model can be learned. We develop *relational structural causal models*, extending structural causal models (Pearl 2009) to settings where objects and their relations vary. First, we show how answers to not only causal but also observational queries about unseen combinations of objects can not be identified without further assumptions. To enable such identification---including in the presence of unobserved confounding---we define *relational causal graphs* and derive symbolic identification criteria. Finally, we propose *relational neural causal models*, a provably correct approach that outperforms non-relational baselines on simulated traffic scenes with varying cars, signals, and pedestrians.

Recently, the spiking neural networks (SNNs) have shown great promise in enhancing AI task performance by utilizing the brain-inspired and energy-efficient computational paradigm via the binary (0/1) spikes. Modern SNNs, especially those based on transformers, often require FPGA accelerators or neuromorphic chips to enable spike-driven computations. However, this domain-specific hardware is not always accessible on commodity edge devices like NVIDIA Jetsons, which may degrade SNNs' energy efficiency due to massive computational waste on inactive "0" spikes and finally undermine the usage boundary. This limitation raises an interesting question: is it possible to make SNNs edge-friendly and tame the computations mostly on active "1" spikes? We present the answer yes and propose Spik4lite, which serves as a lightweight plug-and-play module to significantly improve SNN's performance between model accuracy and computational efficiency. The key is to refactor SNN's channel-wise neuromorphic sparsity by zeroing out low-efficiency channels while proactively compensating for the eliminated spikes. Different from prior methods mainly focusing on optimizing the theoretical synaptic operations, our design philosophy can evolve the SNNs into a physically compact manner, thus inherently saving more computational and energy costs. Extensive experiments based on real edge devices show that Spik4lite can be integrated into existing SNN baselines to further improve their accuracy-and-efficiency performance, guaranteeing the model accuracy while saving the computational and energy costs.

We present MultiBreak, a scalable and diverse multi-turn jailbreak benchmark to evaluate large language model (LLM) safety. Multi-turn jailbreaks mimic natural conversational settings, making them easier to bypass safety-aligned LLM than single-turn jailbreaks. Existing multi-turn benchmarks are limited in size or rely heavily on templates, which restrict their diversity. To address this gap, we unify a wide range of harmful jailbreak intents, and introduce an active learning pipeline for expanding high-quality multi-turn adversarial prompts, where a generator is iteratively fine-tuned to produce stronger attack candidates, guided by uncertainty-based refinement. Our MultiBreak includes 10,389 multi-turn adversarial prompts, spans 2,665 distinct harmful intents, and covers the most diverse set of topics to date. Empirical evaluation shows that our benchmark achieves up to a 54.0% and 34.6% higher attack success rate (ASR) than the second-best dataset on DeepSeek-R1-7B and GPT-4.1-mini, respectively. More importantly, safety evaluations suggest that diverse attack categories uncover fine-grained LLM vulnerabilities, and categories that appear benign under single-turn can exhibit substantially higher adversarial effectiveness in multi-turn scenarios. These findings highlight persistent vulnerabilities of LLMs under realistic adversarial settings and establish MultiBreak as a scalable resource for advancing LLM safety.

End-to-end online map generation is a core component of autonomous driving perception systems. However, in complex traffic scenes, Bird’s-Eye-View (BEV) with vectorized mapping suffers from limitations such as target misclassification, spatial localization drift, and ambiguous semantic segmentation. Introducing uncertainty quantification can alleviate these problems, so we propose MapUQ, a robust BEV vectorized mapping method guided by uncertainty-aware optimization. Specifically, we quantify uncertainty at the feature level to enhance semantic perception, apply an error-driven dynamic receptive field adaptation mechanism at the decoding stage to enforce geometric consistency, and leverage negative sample information at the output head to improve lane classification accuracy. Experimental results on the nuScenes and Argoverse 2 datasets show that our method outperforms prior approaches in AP across three road types, achieving an average improvement of 1.5% over the baseline with marginal computational overhead. In addition, our method surpasses the baseline on uncertainty metrics such as ECE and NLL, significantly improving robustness and mapping accuracy in complex scenarios. Our code has been released at github: https://anonymous.4open.science/r/MapUQ-D287.

Large Reasoning Models (LRMs) enhance performance by generating explicit Chain-of-Thought (CoT) trajectories, yet enabling them to self-evaluate correctness without external supervision remains a critical challenge. Existing methods often rely on ground-truth labels or shallow output probabilities, neglecting the layerwise evolution of the reasoning trajectory. In this work, we introduce GeoR (Geometry of Reasoning), a white-box self-evaluation framework based on layerwise trajectory evolution. GeoR decomposes reasoning fidelity into two complementary dimensions: (1) Geometric Evolution, which synthesizes the first- and second-order evolution of layerwise hidden-state trajectories to quantify geometric progress in reasoning; and (2) Difficulty-Aware Calibration, which utilizes cross-entropy of reasoning progress to normalize the Geometric Evolution against intrinsic query uncertainty. By jointly modeling these factors, GeoR effectively distinguishes the coherent evolution of correct reasoning from the chaotic trajectories of errors. Extensive experiments across eight LRMs and seven benchmarks demonstrate that GeoR consistently outperforms state-of-the-art baselines in AUROC, AUPR, and FPR@95.

Learning latent representations that capture both semantic and spatial information is central to efficient spatio-semantic reasoning. However, many existing approaches rely on implicit latent structures combined with dense feature maps or task-specific heads, limiting computational efficiency and flexibility. We propose WorldComp2D, a novel lightweight representation learning framework that explicitly structures latent space geometry according to object identity and spatial proximity using multiscale \textit{local} receptive fields. This framework consists of (i) a proximity-dependent encoder that maps a given observation into a spatio-semantic latent space and (ii) a localizer that infers the coordinates of objects in the input from the resulting spatio-semantic representation. Using facial landmark localization as a proof-of-concept, we show that, compared to SoTA lightweight models, WorldComp2D reduces the numbers of parameters and FLOPs by up to $4.0\times$ and $2.2\times$, respectively, while maintaining real-time performance on CPU. These results demonstrate that explicitly structured latent spaces provide an efficient and general foundation for spatio-semantic reasoning. This framework is open-sourced at https://github.com/JinSeongmin/WorldComp2D.