NeurIPS 2026 Papers

Text-to-image flow matching transformers degrade sharply in long-tail settings: tail-class outputs collapse in fidelity and diversity, limiting their value as synthetic augmentation for rare conditions. We trace this to low head-versus-tail gradient alignment during fine-tuning, an optimization-level pathology that conditioning- and sampling-side interventions do not address. We propose GRASP (Guided Residual Adapters with Sample-wise Partitioning): a deterministic partition of the conditioning space, paired with group-specific residual adapters in the transformer feedforward layers, that leaves the flow-matching objective and the sampler untouched. In conditional flow matching, condition values index distinct sets of probability paths, so partitioning along the conditioning is the structurally correct factorization suitable as gradient alignment proxy. Because the partition is static, every tail sample is guaranteed to update its assigned expert, which bypasses extreme longtail failure modes. Crucially, GRASP is non-invasive and composable: on MIMIC-CXR-LT, combining GRASP with self-guided minority sampling at inference time yields the best all-labels IRS we observe, beyond either intervention alone. GRASP itself reduces overall FID by up to 80\% and lifts tail-class coverage by up to 44\% over full fine-tuning, learned-routing MoE, and minority guidance. Used as training data for a downstream DenseNet classifier on NIH-CXR-LT, GRASP synthetics significantly outperform every non-GRASP alternative on macro F1, match the macro F1 obtained from real training data, and yield nonzero F1 on $9$ of $13$ classes versus $3$ of $13$ from full fine-tuning. Results on ImageNet-LT confirm the mechanism is not tied to medical inductive bias.

Vision-language models (VLMs) for radiology have emerged as a scalable paradigm by leveraging image-report pairs naturally produced in clinical workflows. However, this pairing reveals a mismatch in scale: each finding occupies only a small region of the image, yet supervision is provided only at the global image-report level. This poses a central challenge: prior approaches spread weight densely across all patches rather than concentrating on the sparse subset relevant to a given query. To address this, we present GLINT (Gated Language-Image alignmeNT), a framework that explicitly models this sparse correspondence. On the alignment side, we introduce Sparsely Gated Alignment, a novel architecture in which a sigmoid gate over a separate gate embedding space activates only the patches relevant to each textual query, enforcing explicit sparsity. On the representation side, we add Dense Feature Regularization, which anchors the trainable encoder's intermediate features to a frozen self-supervised learning (SSL) teacher, preserving the fine-grained patch features that the gate relies on. The same recipe applies to both 2D chest X-ray (CXR) and 3D chest computed tomography (CT), built with DINOv3 and V-JEPA 2.1, respectively. GLINT enables zero-shot classification, grounding, and segmentation from free-text queries, and to our knowledge is the first to demonstrate zero-shot segmentation on 3D CT volumes without mask supervision. Notably, the most pronounced gains arise on zero-shot grounding and segmentation, where sparse, query-specific localization is required, consistent with our design intent. In downstream evaluation, GLINT outperforms both SSL encoders and medical VLMs on classification, report generation, and segmentation.

Although Large Vision-Language Models (LVLMs) have achieved remarkable progress, complex visual reasoning remains challenging. Existing approaches suffer from fundamental limitations: *Thinking about Image* mainly performs reasoning in text space, *Thinking with Image* relies on external tools to introduce additional visual evidence, and *Latent Visual Reasoning*, despite moving reasoning into latent space, still focuses largely on static visual modeling and lacks explicit characterization of temporal dynamics. We therefore propose **Dynamic Latent Visual Reasoning (DLVR)**, a training framework that enables LVLMs to reason about dynamics in continuous latent space from only a static image. We introduce Mahalanobis Novelty Token Selection and Novelty-Adaptive Temporal Quantization to construct dynamic latent supervision from open-source datasets, building DLVR-SFT-100K and DLVR-RL-4K. We further develop a two-stage SFT pipeline that first builds temporal grounding over explicit dynamic processes and then teaches the model to encode dynamic semantics and temporal structure into latent tokens. Finally, we propose Contrastive Dynamic Latent Policy Optimization, which encourages latent trajectories to align with real dynamics while moving away from counterfactual ones. DLVR empirically improves both dynamic-centric and general visual reasoning, achieving an average gain of 10.82% on BabyVision and a 10.50% improvement on HRBench4K, demonstrating the promise of dynamic latent reasoning for LVLMs.

Generating humorous memes is a challenging multimodal task that moves beyond direct image-to-caption supervision. It requires a nuanced reasoning over visual content, contextual cues, and subjective humor. To bridge this gap between visual perception and humorous punchline creation, we propose HUMOR, a novel framework that guides VLMs through hierarchical reasoning and aligns them with group-wise human preferences. First, HUMOR employs a hierarchical, multi-path Chain-of-Thought (CoT): the model begins by identifying a template-level intent, then explores diverse reasoning paths under different contexts, and finally anchors onto a high-quality, context-specific path. This CoT supervision, which traces back from ground-truth captions, enhances reasoning diversity. We further analyze that this multi-path exploration with anchoring maintains a high expected humor quality, under the practical condition that high-quality paths retain significant probability mass. Second, to capture subjective humor, we train a pairwise reward model that operates within groups of memes sharing the same template. Following established theory, this approach ensures a consistent and robust proxy for human preference, even with subjective and noisy labels. The reward model then enables a group-wise RL optimization, providing a theoretical guarantee for monotonic improvement within the trust region. Extensive experiments show that HUMOR empowers various VLMs with superior reasoning diversity, more reliable preference alignment, and higher overall meme quality. Beyond memes, our work presents a general paradigm for open-ended, human-aligned multimodal generation, where success is guided by comparative judgment within coherent output groups.

Discrete diffusion models have emerged as a promising direction for vision-language tasks, offering bidirectional context modeling and theoretical parallelization. However, their practical application is severely hindered by a train-inference discrepancy, which leads to catastrophic error cascades: initial token errors during parallel decoding pollute the generation context, triggering a chain reaction of compounding errors and leading to syntactic errors and semantic hallucinations. To address this fundamental challenge, we reframe the generation process from passive denoising to active refining. We introduce ReDiff, a refining-enhanced diffusion framework that teaches the model to identify and correct its own errors. Our approach features a two-stage training process: first, we instill a foundational revision capability by training the model to revise synthetic errors; second, we implement a novel online self-correction loop where the model is explicitly trained to revise its own flawed drafts by learning from an expert's corrections. This mistake-driven learning endows the model with the crucial ability to revisit and refine its already generated output, effectively breaking the error cascade. Extensive experiments demonstrate that ReDiff significantly improves the coherence and factual accuracy of generated content, enabling stable and efficient parallel generation far superior to traditional denoising methods.

Large Vision-Language Models (LVLMs) have achieved impressive progress in multimodal reasoning, yet they remain prone to object hallucinations, generating descriptions of objects that are not present in the input image. In this work, we investigate hallucination from the perspective of attention-value dynamics inside LVLM vision encoders. We identify a consistent three-phase structure of visual processing---diffusion, focus, and rediffusion---and show that the focus phase is where attention most clearly separates strongly and weakly supported visual tokens. However, low attention does not necessarily imply negligible downstream influence: low-attention tokens in the focus phase can still exert non-negligible value-side influence on the attention output relative to their small attention mass. Through controlled phase-wise value interventions, we find that hallucination behavior is particularly sensitive to the value content of low-attention tokens during this phase. Replacing or neutralizing these values reduces hallucination metrics while largely preserving grounded object evidence. A token-level teacher-forcing analysis further shows that the intervention reduces the probability of hallucinated object tokens with much smaller effects on ground-truth object tokens. In addition, Visual Attention Ratio (VAR) analysis shows that focus-phase intervention is accompanied by increased attention to visual tokens during decoding. Based on these observations, we instantiate a simple training-free inference-time intervention that replaces focus-phase low-attention values with an image-level mean value vector using statistics from a single forward pass. Experiments across multiple LVLM backbones demonstrate that this analysis-derived intervention reduces object hallucination with negligible additional runtime and remains compatible with existing inference-time mitigation methods. Our project page is available at: https://anonymous.4open.science/w/FocusMatters-7B32/

Monocular geometry estimation has recently achieved impressive performance across diverse scenes. However, state-of-the-art models still face notable distortion in local 3D structure, especially in fine details, like thin structure and small objects. We attribute this limitation to an architectural mismatch: most current models decode 3D geometry within a 2D parameterization, where feature interactions are governed by image-plane proximity rather than true 3D spatial relationships. This inadvertently mixes features from geometrically distant surfaces, resulting in over-smoothed geometry particularly around thin or elongated structure. In this paper, we propose a fine-detail monocular geometry estimation with Self-Guided Sparse 3D Refinement (SSR) that lifts monocular geometry modeling from 2D image space to 3D space for high-fidelity metric-scale point map. Our model lifts the coarse point map from a foundation base model onto a sparse voxel shell and refines it via SSR. The SSR employs sparse convolutions that aggregate features based on 3D spatial locality, avoiding feature mixing across depth discontinuities. Extensive experiments on diverse datasets demonstrate that our method significantly outperforms existing approaches in recovering fine detailed 3D geometry across both quantitative metrics and qualitative visualizations.

We present a vision–language pretraining (VLP) framework for the medical domain that focuses on preserving fine-detail clinical information, such as disease severity, which plays a critical role in downstream medical applications. Despite recent advances in medical VLP, existing approaches often struggle to retain such subtle semantics. This limitation mainly stems from the use of contrastive learning schemes that treat supervision as hard one-hot labels, overlooking the ordinal and overlapping nature of clinical findings and thereby introducing false negatives. To overcome these challenges, we propose FinAl, a fine-grained alignment approach that incorporates soft alignment targets into vision-language pretraining. These targets are derived from structured radiology reports and encode relative semantic relationships among clinical descriptions, allowing fine-grained linguistic information to be transferred into the visual representation space. Extensive experiments show that FinAl improves performance on fine-grained clinical tasks, including severity grading and severity-aware retrieval, with additional evaluations on challenging fine-grained scenarios further supporting its robustness. At the same time, the proposed method maintains competitive performance on standard coarse-grained clinical benchmarks.

Federated adaptation of large vision-language models (VLMs) is appealing for privacy-sensitive applications but must contend with client data heterogeneity, strict communication budgets, and on-device inference constraints. Recent methods predominantly introduce additive adaptation interfaces, such as learnable prompts or adapter modules, which attach extra trainable components to the frozen backbone, inflating both parameter counts and inference latency. We challenge this additive paradigm and ask: can robust federated VLM adaptation be achieved through a compact intrinsic interface alone? We answer this question with \textbf{FedIBS} (Federated Intrinsic Bias Selection), a module-free framework that fine-tunes only the biases in feed-forward network projections. FedIBS operates on existing backbone parameters and introduces zero additional inference cost. To make such compact intrinsic adaptation viable under heterogeneous clients, we further propose a Fisher-guided progressive masking mechanism that disentangles shareable and personalized bias coordinates. Each client retains personalized coordinates locally and communicates only the shareable subset for aggregation, enabling sparse communication while preserving client-specific adaptation. Extensive experiments on diverse datasets and heterogeneity settings demonstrate that FedIBS consistently outperforms recent federated prompt- and adapter-based baselines in both personalization and generalization, all while achieving substantially lower communication cost.

We present FM-ChangeNet, a pathwise-supervised framework for change detection that reformulates bi-temporal reasoning as continuous transport in feature space rather than static endpoint comparison. Given encoded pre and post-temporal representations, we construct intermediate latent states and learn a time-conditioned velocity field $\hat{v}_\theta(z_t,t)$ along the transformation trajectory. This pathwise formulation constrains the predictor over a continuum of intermediate states, providing a denser and less ambiguous supervision signal than conventional endpoint-only segmentation and enabling the model to capture temporal evolution explicitly. The learned velocity field is not only a transport mechanism but also an interpretable representation of change: its magnitude serves as a spatially localized change cue that helps distinguish true structural variation from nuisance effects such as illumination shifts and spatial misalignment. We develop a hierarchical multi-scale architecture with cross-temporal alignment, time-conditioned coarse-to-fine flow decoding, and a unified objective that couples flow supervision, trajectory consistency, spatial regularization, and segmentation loss. Experiments on remote sensing benchmarks show that the proposed framework produces more structured and robust change representations while achieving state-of-the-art performance.

Three-dimensional (3D) anomaly detection underpins quality assurance in advanced manufacturing and engineering, capturing subtle defects and deviations that 2D inspection struggles to resolve due to occlusion, viewpoint, and appearance confounds. Existing 3D anomaly detection methods either rely on reconstruction errors or stored normal representations to detect anomalies. While both paradigms have driven substantial progress, they share a fundamental limitation: both operate entirely in the visual domain, missing the rich semantics encoded in natural language, which also typically constrains them to per-category models. We propose FACETS, among the first frameworks that leverage a 3D vision--language model for 3D anomaly detection, opening a new paradigm beyond reconstruction- and memory-bank-based methods. The key idea is to explicitly retain native point-level features that enable reasoning at both patch and point granularities, and further leverage language grounding and cross-granularity geometric modeling along two complementary axes, linguistic semantics and geometric saliency, so that coarse semantic cues and fine-grained geometric details jointly support anomaly detection. FACETS enables unified multi-category anomaly detection, avoiding the per-category models used by most prior methods. Notably, we provide a mathematical analysis of our loss function, offering valuable insights into the substantial improvement FACETS achieves in anomaly localization. Extensive experiments on popular benchmarks reveal that FACETS establishes new SOTA performance across datasets and metrics for 3D anomaly detection, substantially and consistently outperforming existing methods. Code is provided as supplementary material.

Feed-forward 3D reconstruction from unconstrained multi-view images has recently emerged as a dominant paradigm in 3D computer vision. Various foundational reconstruction models have been proposed with different architectures and reconstruction paradigms. Despite being trained on large-scale perspective datasets with substantial computational resources, these models generally fail when deployed on images whose camera models deviate from the training distribution. Applying such models to downstream applications with arbitrary, unknown, and potentially mixed camera types remains challenging. To maximally leverage the strengths of existing reconstruction models with minimal retraining efforts, we propose a low-rank adaptation framework that efficiently extends perspective-pretrained models to fisheye and panoramic imagery. Extensive experiments demonstrate that, with only 0.1% additional parameters and one day of training, our method can adapt a foundational 3D reconstruction model to non-pinhole cameras while fully preserving its original performance on perspective inputs.

Combining the Segment Anything Model (SAM) with fine-tuning techniques allows SAM to be effectively adapted to various downstream image segmentation tasks. However, this adaptability introduces new security vulnerabilities related to adversarial attacks. In this paper, we investigate the adversarial transferability between the original SAM and its fine-tuned downstream models. Under limited knowledge conditions of the downstream models, we propose a novel structure-exploiting transferable attack (SETA) method. Our framework mimics the fine-tuning architecture and estimates the parameter distributions of the downstream models to improve the transferability of the generated adversarial samples. Experimental results demonstrate the efficacy of our proposed method in creating adversarial examples against various downstream fine-tuned SAM models.

The development of native computer-use agents (CUA) represents a significant leap in multimodal AI. However, their potential is currently bottlenecked by the constraints of static data scaling. Existing paradigms relying primarily on passive imitation of static datasets struggle to capture the intricate causal dynamics inherent in long-horizon computer tasks. In this work, we introduce EvoCUA, a native agentic model that integrates data generation and policy optimization into a self-sustaining evolutionary cycle. This approach employs a verifiable synthesis engine to autonomously generate diverse tasks with executable validators, alongside a scalable infrastructure orchestrating tens of thousands of sandbox rollouts for mass experience acquisition. To internalize this experience, our iterative evolving learning strategy reinforces successful routines while transforming failure trajectories into rich supervision through error analysis and self-correction.EvoCUA achieves a 56.7\% success rate on the OSWorld benchmark, establishing a new state-of-the-art for open-weights models across multiple model scales. It significantly outperforms the previous best open-source model, OpenCUA-72B (45.0%), and surpasses leading closed-weights models such as UI-TARS-2 (53.1%), offering a highly parameter-efficient and reproducible alternative to massive closed-weights models. These results demonstrate the generalizability of the evolving paradigm across foundation models of varying scales, establishing a robust and scalable path for advancing native agent capabilities.

Vision-Language-Action (VLA) models have emerged as a promising paradigm for robotic manipulation by unifying perception, language grounding, and action generation. However, they often struggle in scenarios requiring precise spatial understanding, as current VLA models primarily rely on 2D visual representations that lack depth information and detailed spatial relationships. While recent approaches incorporate explicit 3D inputs such as depth maps or point clouds to address this issue, they often increase system complexity, require additional sensors, and remain vulnerable to sensing noise and reconstruction errors. Another line of work explores implicit 3D-aware spatial modeling directly from RGB observations without extra sensors, but it often relies on large geometry foundation models, resulting in higher training and deployment costs. To address these challenges, we propose \textbf{Evo-Depth}, a lightweight depth-enhanced VLA framework that enhances spatially grounded manipulation without relying on additional sensing hardware or compromising deployment efficiency. Evo-Depth employs a lightweight \textit{Implicit Depth Encoding Module} (IDEM) to extract compact depth features from multi-view RGB images. These features are incorporated into vision-language representations through a \textit{Spatial Enhancement Module} (SEM) via depth-aware modulation, enabling efficient spatial-semantic enhancement. A \textit{Progressive Alignment Training} strategy is further introduced to align the resulting depth-enhanced representations with downstream action learning. Extensive experiments in simulation and real-world settings demonstrate the effectiveness of Evo-Depth. With only \textbf{0.9B} parameters, Evo-Depth achieves state-of-the-art-level performance across four simulation benchmarks. In real-world experiments, Evo-Depth attains the highest average success rate while also exhibiting the smallest model size, lowest GPU memory usage, and highest inference frequency among compared methods. These results demonstrate that lightweight implicit depth enhancement is an effective and practical solution for spatially grounded robotic manipulation. We release code and models to facilitate future research on lightweight depth-enhanced VLA models.

Vision-Language Models (VLMs) have achieved strong performance on static visual understanding benchmarks, yet their ability to act in dynamic environments remains poorly understood. We introduce AtariBench, a controlled benchmark that evaluates VLMs across 20 Atari 2600 games, requiring models to make sequential decisions from high-temporal-resolution gameplay streams. By decoupling fast in-game dynamics from wall-clock inference speed, AtariBench focuses on whether models can perceive rapid visual state changes, reason over spatiotemporal dynamics, and ground actions in the current observation. Our experiments reveal substantial limitations in state-of-the-art VLMs: models often fail to infer object motion and direction, exhibit brittle sequential decision-making, and increasingly over-rely on textual interaction history rather than newly appended visual feedback. We further introduce a separability analysis to identify which in-game scores provide reliable signals of model competence, and use the benchmark to study how memory design and prompting strategies affect performance. Overall, AtariBench exposes a persistent gap between passive visual perception and active decision-making, providing a lightweight and rigorous testbed for developing VLMs that can reason and act in dynamic environments.

Gaze following requires both scene understanding and gaze reasoning to localize the gaze target of an in-scene person. Recently, vision foundation models (VFMs) have demonstrated strong performance on this task, enabling simpler architectures while outperforming prior methods. However, we observe a key limitation of VFM-based approaches: while VFMs substantially improve scene understanding, they contribute little to gaze reasoning. As a result, existing methods often rely on semantically salient objects rather than true gaze cues, leading to degraded performance when targets are not salient. To address this, we propose a novel training mechanism to enhance gaze reasoning in VFMs for gaze following. Our method includes: (1) a head-conditioned local LoRA, which enables localized adaptation to preserve scene token learning while improving head token learning for gaze reasoning; and (2) an out-of-cone penalty, which injects gaze cues into head tokens while aligning them with scene tokens. Experiments on the GazeFollow and VAT datasets demonstrate that our method achieves state-of-the-art performance, with particularly strong improvements when gaze targets are not semantically salient. Our findings offer valuable insights for advancing future gaze following research. We will release the code once the paper is accepted.

Vision-Language-Action (VLA) models have emerged as a powerful paradigm for generalist robotic control. However, their high computational cost and limited control frequency hinder real-time robotic manipulation, especially when large vision-language backbones and iterative action heads are executed at every control step. Existing VLA acceleration methods often optimize individual components or rely on fixed acceleration rules, treating different control steps with largely fixed computation and overlooking the non-uniform reasoning demands of sequential embodied control. Inspired by human motor control, where cognitive and feedback resources concentrate on goal-sensitive stages, we argue that VLA models should learn when to invest full computation and when to reuse prior computation. To this end, we propose ElegantVLA, a plug-in phase-adaptive inference framework that accelerates VLA models through intra-model dynamic compute scheduling. ElegantVLA introduces a lightweight scheduler that observes temporal representation similarity, robot-motion cues, and episode progress to jointly allocate computation across the vision encoder, LLM, and action head. For perception-language reasoning, the scheduler selects a five-level Vision--LLM compute mode, from full recomputation to multi-step temporal reuse, based on visual-language representation stability. For action generation, it selects a three-level denoising mode, reusing intermediate denoising states during stable motion while preserving full refinement for goal-sensitive stages. By coordinating these decisions, ElegantVLA provides a general acceleration framework for modern VLA pipelines with explicit action-generation modules, without modifying or retraining the base model. Extensive experiments on GR00T, CogACT, and real-world tasks show that ElegantVLA preserves or improves task success while substantially accelerating inference. On GR00T, it achieves up to 2.55$\times$ average speedup. On CogACT, it delivers a 3.77$\times$ average speedup. In GR00T-based real-world experiments across six tasks, it reduces computation by 2.18$\times$ and increases control frequency from 13.8 Hz to 26.3 Hz.

Touch is fundamental to dexterous manipulation, yet most egocentric human data increasingly used for robot learning lacks tactile information. Directly collecting large-scale tactile data is challenging due to sensor limitations, while human video data is abundant, contact-rich, and easily scalable. This motivates a natural question: can tactile signals be inferred purely from vision? To address this, we introduce **EgoTac**, a generalizable model that predicts rich **tac**tile information directly from **ego**centric human videos. EgoTac is trained on a unified corpus of over 5.7M image-tactile pairs, covering both continuous force measurements and binary contacts. By learning from this diverse dataset, EgoTac captures nuanced touch dynamics across varied interactions. Experiments demonstrate strong performance: in-domain prediction achieves an average force error below 0.06N. On out-of-domain contact prediction benchmarks, EgoTac consistently outperforms the state-of-the-art contact estimator. It also captures the rise and fall patterns of real tactile data and enables zero-shot predictions on unconstrained real-world videos. Scaling analyses further reveal that both data diversity and volume improve performance steadily. Overall, EgoTac provides a scalable pathway to extract tactile priors from egocentric human videos, enabling broadly applicable tactile-aware robot learning.