InsightTok: Improving Text and Face Fidelity in Discrete Tokenization for Autoregressive Image Generation

Text and faces are among the most perceptually salient and practically important patterns in visual generation, yet they remain challenging for autoregressive generators built on discrete tokenization. A central bottleneck is the tokenizer: aggressive downsampling and quantization often discard the fine-grained structures needed to preserve readable glyphs and distinctive facial features. We attribute this gap to standard discrete-tokenizer objectives being weakly aligned with text legibility and facial fidelity, as these objectives typically optimize generic reconstruction while compressing diverse content uniformly. To address this, we propose InsightTok, a simple yet effective discrete visual tokenization framework that enhances text and face fidelity through localized, content-aware perceptual losses. With a compact 16k codebook and a 16x downsampling rate, InsightTok significantly outperforms prior tokenizers in text and face reconstruction without compromising general reconstruction quality. These gains consistently transfer to autoregressive image generation in InsightAR, producing images with clearer text and more faithful facial details. Overall, our results highlight the potential of specialized supervision in tokenizer training for advancing discrete image generation.

Steering Visual Generation in Unified Multimodal Models with Understanding Supervision

Unified multimodal models are envisioned to bridge the gap between understanding and generation. Yet, to achieve competitive performance, state-of-the-art models adopt largely decoupled understanding and generation components. This design, while effective for individual tasks, weakens the connection required for mutual enhancement, leaving the potential synergy empirically uncertain. We propose to explicitly restore this synergy by introducing Understanding-Oriented Post-Training (UNO), a lightweight framework that treats understanding not only as a distinct task, but also a direct supervisory signal to steer generative representations. By incorporating objectives that encode semantic abstraction (captioning) and structural details (visual regression), we enable effective gradient flow from understanding to generation. Extensive experiments on image generation and editing demonstrate that understanding can serve as an effective catalyst for generation.

Linear-Time Global Visual Modeling without Explicit Attention

Existing research largely attributes the global sequence modeling capability of Transformers to the explicit computation of attention weights, a process that inherently incurs quadratic computational complexity. In this work, we offer a novel perspective: we demonstrate that attention can be mathematically reframed as a Multi-Layer Perceptron (MLP) equipped with dynamically predicted parameters. Through this lens, we explain attention's global modeling power not as explicit token-wise aggregation, but as an implicit process where dynamically generated parameters act as a compressed representation of the global context. Inspired by this insight, we investigate a fundamental question: can we achieve Transformer-level sequence global modeling entirely through dynamic parameterization while maintaining linear complexity, effectively replacing explicit attention? To explore this, we design various dynamic parameter prediction strategies and integrate them into standard network layers. Extensive empirical studies on vision models demonstrate that dynamic parameterization can indeed serve as a highly effective, linear-complexity alternative to explicit attention, opening new pathways for efficient sequence modeling. Code is available at https://github.com/LeapLabTHU/WeightFormer.

Linearizing Vision Transformer with Test-Time Training

While linear-complexity attention mechanisms offer a promising alternative to Softmax attention for overcoming the quadratic bottleneck, training such models from scratch remains prohibitively expensive. Inheriting weights from pretrained Transformers provides an appealing shortcut, yet the fundamental representational gap between Softmax and linear attention prevents effective weight transfer. In this work, we address this conversion challenge from two perspectives: architectural alignment and representational alignment. We identify Test-Time Training (TTT) as a linear-complexity architecture whose two-layer dynamic formulation is structurally aligned with Softmax attention, enabling direct inheritance of pretrained attention weights. To further align representational properties, including key shift-invariance and locality, we introduce key instance normalization and a lightweight locality enhancement module. We validate our approach by linearizing Stable Diffusion 3.5 and introduce SD3.5-T$^5$ (Transformer To Test Time Training). With only 1 hour of fine-tuning on 4$\times$H20 GPUs, SD3.5-T$^5$ achieves comparable text-to-image quality to the fine-tuned Softmax model, while accelerating inference by 1.32$\times$ and 1.47$\times$ at 1K and 2K resolutions.

Refinement via Regeneration: Enlarging Modification Space Boosts Image Refinement in Unified Multimodal Models

Unified multimodal models (UMMs) integrate visual understanding and generation within a single framework. For text-to-image (T2I) tasks, this unified capability allows UMMs to refine outputs after their initial generation, potentially extending the performance upper bound. Current UMM-based refinement methods primarily follow a refinement-via-editing (RvE) paradigm, where UMMs produce editing instructions to modify misaligned regions while preserving aligned content. However, editing instructions often describe prompt-image misalignment only coarsely, leading to incomplete refinement. Moreover, pixel-level preservation, though necessary for editing, unnecessarily restricts the effective modification space for refinement. To address these limitations, we propose Refinement via Regeneration (RvR), a novel framework that reformulates refinement as conditional image regeneration rather than editing. Instead of relying on editing instructions and enforcing strict content preservation, RvR regenerates images conditioned on the target prompt and the semantic tokens of the initial image, enabling more complete semantic alignment with a larger modification space. Extensive experiments demonstrate the effectiveness of RvR, improving Geneval from 0.78 to 0.91, DPGBench from 84.02 to 87.21, and UniGenBench++ from 61.53 to 77.41.

Bridging the RGB-IR Gap: Consensus and Discrepancy Modeling for Text-Guided Multispectral Detection

Text-guided multispectral object detection uses text semantics to guide semantic-aware cross-modal interaction between RGB and IR for more robust perception. However, notable limitations remain: (1) existing methods often use text only as an auxiliary semantic enhancement signal, without exploiting its guiding role to bridge the inherent granularity asymmetry between RGB and IR; and (2) conventional data-driven attention-based fusion tends to emphasize stable consensus while overlooking potentially valuable cross-modal discrepancies. To address these issues, we propose a semantic bridge fusion framework with bi-support modeling for multispectral object detection. Specifically, text is used as a shared semantic bridge to align RGB and IR responses under a unified category condition, while the recalibrated thermal semantic prior is projected onto the RGB branch for semantic-level mapping fusion. We further formulate RGB-IR interaction evidence into the regular consensus support and the complementary discrepancy support that contains potentially discriminative cues, and introduce them into fusion via dynamic recalibration as a structured inductive bias. In addition, we design a bidirectional semantic alignment module for closed-loop vision-text guidance enhancement. Extensive experiments demonstrate the effectiveness of the proposed fusion framework and its superior detection performance on multispectral benchmarks. Code is available at https://github.com/zhenwang5372/Bridging-RGB-IR-Gap.

MAG-3D: Multi-Agent Grounded Reasoning for 3D Understanding

Vision-language models (VLMs) have achieved strong performance in multimodal understanding and reasoning, yet grounded reasoning in 3D scenes remains underexplored. Effective 3D reasoning hinges on accurate grounding: to answer open-ended queries, a model must first identify query-relevant objects and regions in a complex scene, and then reason about their spatial and geometric relationships. Recent approaches have demonstrated strong potential for grounded 3D reasoning. However, they often rely on in-domain tuning or hand-crafted reasoning pipelines, which limit their flexibility and zero-shot generalization to novel environments. In this work, we present MAG-3D, a training-free multi-agent framework for grounded 3D reasoning with off-the-shelf VLMs. Instead of relying on task-specific training or fixed reasoning procedures, MAG-3D dynamically coordinates expert agents to address the key challenges of 3D reasoning. Specifically, we propose a planning agent that decomposes the task and orchestrates the overall reasoning process, a grounding agent that performs free-form 3D grounding and relevant frame retrieval from extensive 3D scene observations, and a coding agent that conducts flexible geometric reasoning and explicit verification through executable programs. This multi-agent collaborative design enables flexible training-free 3D grounded reasoning across diverse scenes and achieves state-of-the-art performance on challenging benchmarks.

HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning

Vision-language models (VLMs) show strong multimodal capabilities but still struggle with fine-grained vision-language reasoning. We find that long chain-of-thought (CoT) reasoning exposes diverse failure modes, including perception, reasoning, knowledge, and hallucination errors, which can compound across intermediate steps. However, most existing vision-language data used for reinforcement learning with verifiable rewards (RLVR) does not involve complex reasoning chains that rely on visual evidence throughout, leaving these weaknesses largely unexposed. We therefore propose HopChain, a scalable framework for synthesizing multi-hop vision-language reasoning data for RLVR training of VLMs. Each synthesized multi-hop query forms a logically dependent chain of instance-grounded hops, where earlier hops establish the instances, sets, or conditions needed for later hops, while the final answer remains a specific, unambiguous number suitable for verifiable rewards. We train Qwen3.5-35B-A3B and Qwen3.5-397B-A17B under two RLVR settings: the original data alone, and the original data plus HopChain's multi-hop data, and compare them across 24 benchmarks spanning STEM and Puzzle, General VQA, Text Recognition and Document Understanding, and Video Understanding. Although this multi-hop data is not synthesized for any specific benchmark, it improves 20 of 24 benchmarks on both models, indicating broad and generalizable gains. Consistently, replacing full chained queries with half-multi-hop or single-hop variants reduces the average score across five representative benchmarks from 70.4 to 66.7 and 64.3, respectively. Notably, multi-hop gains peak in long-CoT vision-language reasoning, exceeding 50 points in the ultra-long-CoT regime. These experiments establish HopChain as an effective, scalable framework for synthesizing multi-hop data that improves generalizable vision-language reasoning.

UltraStar: Semantic-Aware Star Graph Modeling for Echocardiography Navigation

Echocardiography is critical for diagnosing cardiovascular diseases, yet the shortage of skilled sonographers hinders timely patient care, due to high operational difficulties. Consequently, research on automated probe navigation has significant clinical potential. To achieve robust navigation, it is essential to leverage historical scanning information, mimicking how experts rely on past feedback to adjust subsequent maneuvers. Practical scanning data collected from sonographers typically consists of noisy trajectories inherently generated through trial-and-error exploration. However, existing methods typically model this history as a sequential chain, forcing models to overfit these noisy paths, leading to performance degradation on long sequences. In this paper, we propose UltraStar, which reformulates probe navigation from path regression to anchor-based global localization. By establishing a Star Graph, UltraStar treats historical keyframes as spatial anchors connected directly to the current view, explicitly modeling geometric constraints for precise positioning. We further enhance the Star Graph with a semantic-aware sampling strategy that actively selects the representative landmarks from massive history logs, reducing redundancy for accurate anchoring. Extensive experiments on a dataset with over 1.31 million samples demonstrate that UltraStar outperforms baselines and scales better with longer input lengths, revealing a more effective topology for history modeling under noisy exploration.

TwinRL-VLA: Digital Twin-Driven Reinforcement Learning for Real-World Robotic Manipulation

Despite strong generalization capabilities, Vision-Language-Action (VLA) models remain constrained by the high cost of expert demonstrations and insufficient real-world interaction. While online reinforcement learning (RL) has shown promise in improving general foundation models, applying RL to VLA manipulation in real-world settings is still hindered by low exploration efficiency and a restricted exploration space. Through systematic real-world experiments, we observe that the effective exploration space of online RL is closely tied to the data distribution of supervised fine-tuning (SFT). Motivated by this observation, we propose TwinRL, a digital twin-real-world collaborative RL framework designed to scale and guide exploration for VLA models. First, a high-fidelity digital twin is efficiently reconstructed from smartphone-captured scenes, enabling realistic bidirectional transfer between real and simulated environments. During the SFT warm-up stage, we introduce an exploration space expansion strategy using digital twins to broaden the support of the data trajectory distribution. Building on this enhanced initialization, we propose a sim-to-real guided exploration strategy to further accelerate online RL. Specifically, TwinRL performs efficient and parallel online RL in the digital twin prior to deployment, effectively bridging the gap between offline and online training stages. Subsequently, we exploit efficient digital twin sampling to identify failure-prone yet informative configurations, which are used to guide targeted human-in-the-loop rollouts on the real robot. In our experiments, TwinRL approaches 100% success in both in-distribution regions covered by real-world demonstrations and out-of-distribution regions, delivering at least a 30% speedup over prior real-world RL methods and requiring only about 20 minutes on average across four tasks.