arXiv subjects

Ting Huang

Publications and source records attributed to Ting Huang.

At least 19 recordsLinked to original sources

arXiv2026

RFPO: Rectified Flow Policy Optimization for Embodied Control

Flow-based policies provide an expressive framework for continuous robot control, but their iterative ODE integration incurs substantial inference cost. Naively reducing the integration budget can severely degrade control, since policies optimized under full-step execution are not explicitly constrained to remain reliable under coarse numerical integration. We refer to this mismatch as the few-step discretization gap. To address this problem, we introduce RFPO, a flow-policy optimization framework for reliable few-step execution. Reward-aware online Reflow rectifies student-induced transport paths during on-policy learning, making the resulting policy more robust to coarse integration. A frozen Gaussian PPO controller supplies complementary action-space supervision at full and intermediate integration budgets, while the deployed policy remains a single flow student executed with one Euler step. Across Unitree Go2, Boston Dynamics Spot, Unitree H1, and Unitree G1, RFPO consistently preserves full-step control performance under one-step execution, with one-step returns remaining within 2.4% of their corresponding 64-step values across both zero and random initialization. On Unitree Go2, one-step execution retains 98.5% of the 64-step reward while reducing onboard mean inference latency from 4.39 ms to 0.08 ms, yielding a 54.9x speedup. Real-robot experiments further validate stable one-step locomotion. Code: https://github.com/AIGeeksGroup/RFPO. Website: https://aigeeksgroup.github.io/RFPO.

cs.RO↗

arXiv2026

Mingbird: A Local-First Agent Harness Enabling Small Open Models to Complete Real Tasks

Small open-weight models (2-9B) run on ordinary laptops, but under cloud-scale agent harnesses they rarely complete real tasks: tool prefill overflows the context, self-correction diverges, tool demonstrations loop, and tasks are silently abandoned. We present evidence, from a controlled single-machine comparison and one third-party benchmark, that a substantial share of these failures is attributable to the harness rather than the model. We introduce Mingbird, a local-first agent harness for Windows and Ollama whose ten mechanisms compensate point-by-point for small-model failure forms, three of them representative: a byte-level net-zero prefill budget, a finish gate that re-reads the task before accepting completion, and signature-level loop detection. On LRAB, a controlled comparison holding machine, models, budgets, and scoring fixed (4 harnesses $\times$ 4 open models (2B-35B) $\times$ 18 real tasks, deterministic artifact scoring), Mingbird reaches 0.886 overall against 0.631 (goose), 0.479 (opencode), and 0.405 (agent-mini), with all 288 cells published; on $τ^2$-bench (278 tasks, three arms, one protocol) it totals 0.856 against 0.791 and 0.737; and a frontier-model probe on the same 18 tasks spans 0.997 to 0.478 across harnesses, with well-formed scaffolds staying within 0.072 of each other. A leave-one-mechanism-out ablation is reported as directional only: same-night replications of the same arm move its mean by up to 0.069, the size of every nominal single-trial delta, and the one batch-matched comparison (full mechanism stack versus text re-read alone) gives the executable completion guards a paired +0.10 across three replications. The evidence carries stated limits: a self-built benchmark, a single machine, and single-trial scoring.

cs.AI↗

arXiv2026

Beyond Pixel Reconstruction: Retrieval-Guided Glyph-Aware Restoration for Low-Resource Manchu Historical Documents

Historical Manchu documents preserve invaluable linguistic and cultural heritage, yet their digitization is hindered by severe degradations and the scarcity of paired training data. Existing document restoration methods primarily optimize pixel-level reconstruction, which can produce visually plausible results while failing to preserve the structural identity of Manchu glyphs. To address this limitation, we propose a retrieval-guided glyph-aware restoration framework that goes beyond pixel reconstruction by explicitly incorporating glyph-level structural knowledge. Our method retrieves relevant glyph exemplars to provide structural guidance during restoration and integrates this information into the reconstruction process, improving the recovery of degraded character structures under low-resource conditions. Extensive experiments on Manchu historical documents demonstrate that the proposed approach improves both image restoration quality and glyph-level fidelity compared with existing restoration methods. These results highlight the importance of incorporating character-aware structural priors for reliable restoration of low-resource historical documents.

cs.CV↗

arXiv2026

Think Before You Restore: Risk-Aware Manchu Manuscript Restoration with Stroke-Guided Attention

Full-page blind restoration of historical Manchu manuscripts is challenging due to scarce annotations, unknown degradation regions, and fragile connected strokes. Generic restoration models may improve visual quality but often modify intact content, leading to over-restoration. We propose SAGE-Restore (Stroke-Aware Gated rEstoration), a selective restoration framework that first assesses where restoration is needed and then uses this assessment to guide restoration candidate generation and pixel-level selection. Its encoder predicts patch-level repair probabilities from complementary appearance and stroke-structural cues to condition restoration candidate generation, while the corresponding repair logits are refined into a pixel-level soft gate that selectively controls where the restoration candidate is applied. We further introduce a fidelity-aware evaluation protocol that jointly measures degraded-region recovery, intact-content preservation, and their balance. SAGE-Restore achieves the highest R-Recovery (0.463) and RFS (0.626), while maintaining high U-Fidelity (0.968), demonstrating an effective balance between restoration and content preservation.

cs.CV↗

arXiv2026

Ramsey numbers of sparse graphs versus disjoint books

Let $B_k$ denote a book on $k+2$ vertices and $tB_k$ be $t$ vertex-disjoint $B_k$'s. For any integers $k\ge2$ and $t\ge1$, there exists a positive constant $ε=ε(k,t)$ such that every connected graph $G$ with $n\ge111t^3k^3$ vertices and at most $n(1+ε)$ edges satisfies $$r(G,tB_k)=2n+t-2.$$ Our result extends the work of Erdős, Faudree, Rousseau, and Schelp (1988), who established the corresponding result for $G$ being a tree and $t=1$.

math.CO↗

arXiv2026

Fan-goodness of sparse graphs

Let $G$ be a connected graph of order $n$, $F_k$ be a fan consisting of $k$ triangles sharing a common vertex, and $tF_k$ be $t$ vertex-disjoint copies of $F_k$. Brennan (2017) showed the Ramsey number $r(G,F_k)=2n-1$ for $G$ being a unicyclic graph for $n \geq k^2-k+1$ and $k\ge 18$, and asked the threshold $c(n)$ for which $r(G,F_k) \geq 2n$ holds for any $G$ containing at least $c(n)$ cycles and $n$ being large. In this paper, we consider fan-goodness of general sparse graphs and show that if $G$ has at most $n(1+ε(k))$ edges, where $ε(k)$ is a constant depending on $k$, then $$r(G,F_k)=2n-1$$ for $n\ge 36k^4$, which implies that $c(n)$ is greater than $ε(k) n$. Moreover, if $G$ has at most $n(1+ε(k,t))$ edges, where $ε(k,t)$ is a constant depending on $k,t$, then $$r(G,tF_k)=2n+t-2$$ provided $n\ge 161t^2k^4$.

math.CO↗

arXiv2026

Ramsey multiplicity and extremal colorings for odd cycles

The Ramsey number $r(H)$ of a graph $H$ is the minimum positive integer $N$ such that every red/blue edge-coloring of the complete graph $K_N$ on $N$ vertices contains a monochromatic copy of $H$. The Ramsey multiplicity $M(H,n)$ is the minimum number of monochromatic copies of $H$ over all red/blue edge-colorings of $K_n$. It is called threshold Ramsey multiplicity if $n=r(H)$, and denoted by $m(H)$. The only previously known general infinite family for which $m(H)$ has been determined is stars, due to Harary and Prins (1974). Let $C_k$ denote a cycle on $k$ vertices. Conlon, Fox, Sudakov, and Wei (2022) conjectured that $m(C_k)=(k-1)!/2$ for every sufficiently large odd integer $k$. In this paper, we determine $M(C_k,r(C_k)+\ell)$ for every fixed nonnegative integer $\ell$ and all sufficiently large odd $k$, and characterize all extremal colorings, thereby confirming the conjecture. This is also a second general infinite family for which $m(H)$ has been determined.

math.CO↗

arXiv2026

Minimum degree and sparse connected spanning subgraphs

Let $G$ be a connected graph on $n$ vertices and at most $n(1+ε)$ edges with bounded maximum degree, and $F$ a graph on $n$ vertices with minimum degree at least $n-k$, where $ε$ is a constant depending on $k$. In this paper, we prove that $F$ contains $G$ as a spanning subgraph provided $n\ge 6k^3$, by establishing tight bounds for the Ramsey number $r(G,K_{1,k})$, where $K_{1,k}$ is a star on $k+1$ vertices. Our result generalizes and refines the work of Erdős, Faudree, Rousseau, and Schelp (JCT-B, 1982), who established the corresponding result for $G$ being a tree. Moreover, the tight bound for $r(G,tK_{1,k})$ is also obtained.

math.CO↗

arXiv2026

DramaAgent: Agentic Storytelling Video Generation

Recent diffusion and autoregressive models have substantially improved text-to-video generation, yet producing coherent long-form story videos with consistent characters and aligned audio remains challenging. Existing methods often suffer from narrative drift, unstable character identity, weak cross-scene continuity, and audio-visual mismatch over extended sequences. We propose DramaAgent, a hierarchical, agentic, and model-agnostic framework for long-form text-to-video-and-audio generation. Rather than improving the underlying video backbone itself, DramaAgent introduces an upper-level control layer that decomposes generation into story planning, persistent character conditioning, scene-wise synthesis, and reflection-guided targeted repair. The framework maintains reusable story and character states across scenes, diagnoses failures such as identity drift, missing scene semantics, temporal discontinuity, and cross-modal mismatch, and repairs problematic clips in a stage-specific manner. Experiments across multiple video generation backbones show that DramaAgent improves long-horizon coherence, character consistency, narrative fidelity, and scene-level audio-visual consistency over direct generation and strong baselines. These results suggest that hierarchical agentic control is a practical direction for controllable long-form audiovisual generation. Code: https://github.com/AIGeeksGroup/DramaAgent. Website: https://aigeeksgroup.github.io/DramaAgent.

cs.CV↗

arXiv2026

MobileVLA-R1 2.0: RL-Enhanced Reasoning for Mobile Robot Control

Grounding natural-language instructions into reliable and executable actions remains a fundamental challenge for vision-language-action (VLA) systems on mobile robots, due to the persistent gap between high-level semantic reasoning and low-level locomotion and manipulation control. Existing approaches often rely on implicit reasoning or monolithic action prediction, making it difficult to maintain coherent long-horizon decision making while producing precise and adaptable robot actions. To address this challenge, we propose MobileVLA-R1 2.0, an RL-enhanced VLA framework that explicitly couples structured embodied reasoning with executable mobile robot control. The framework learns multi-granularity reasoning over embodied trajectories through supervised Chain-of-Thought (CoT) alignment and reinforcement learning, improving reasoning-to-action consistency beyond purely behavioral supervision. To support both locomotion and manipulation, we further introduce a reasoning-conditioned action decoder that maps multimodal reasoning representations to task-level action targets, which are subsequently translated into embodiment-specific commands by robot controllers. This design provides a unified perception-reasoning-action interface while decoupling high-level action generation from robot-specific actuation. We conduct extensive evaluations on language-guided navigation, quadruped control, and humanoid mobile manipulation, covering VLN-CE, QUARD, and real-world deployments on Unitree Go2 and G1 robots. MobileVLA-R1 2.0 consistently outperforms strong VLA baselines, achieving an average 1.6 point improvement in SR on VLN-CE and a 10.0 point improvement in full-task success on real-world G1 mobile manipulation tasks over MobileVLA-R1, while demonstrating robust long-horizon instruction following and closed-loop execution across different robotic platforms.

cs.CV↗

arXiv2026

REER-PT: Reverse-Engineered Reasoning for Perplexity-Guided Pre-training Data Augmentation

As language-model compute continues to scale, high-quality training data is becoming an increasingly important bottleneck. Conventional next-token prediction supervises what follows a context but leaves the intermediate reasoning behind that continuation implicit. We introduce \textbf{REER-PT}, a scalable framework that extends Reverse-Engineered Reasoning (REER) to raw pre-training data. REER-PT identifies continuations that are difficult to predict but can still be inferred from the preceding context, and inserts concise reasoning annotations that reconstruct the missing connection between context and continuation. Candidate annotations are generated and refined offline, with perplexity serving as the optimization signal. Constraints on length and target leakage filter out unhelpful or trivial annotations. This sparse transformation preserves the source text and remains compatible with standard next-token prediction, avoiding online reasoning rollouts during pre-training. We apply REER-PT to transform a source pre-training corpus into an augmented one. Across augmented-data, original-token, and selected-continuation comparisons, perplexity reductions range from 0.42 to 7.29, and only about 0.05\% of annotation 13-grams appear verbatim in the source text. We then train two 680M-parameter models with the same architecture and training configuration on the source and augmented corpora, respectively. The augmented-data model gains up to 2.07 percentage points on several knowledge and reasoning benchmarks. Together, the perplexity analysis indicates improved continuation predictability, while the controlled pre-training experiments suggest that this augmentation can improve model performance without changing the standard pre-training objective.

cs.CL↗

arXiv2026

Let Geometry GUIDE: Layer-wise Unrolling of Geometric Priors in Multimodal LLMs

Multimodal Large Language Models (MLLMs) have achieved remarkable progress in 2D visual tasks but still struggle to understand physical space in real-world visual streams. Recently, feed-forward geometric foundation models that implicitly extract geometric priors from visual inputs have offered a promising direction. However, many existing geometry-aware MLLMs extract features from a single deep encoder layer and perform one-shot fusion at the language-model input, potentially overlooking complementary information across geometric encoder layers, limiting access to fine-grained geometric cues, and hindering progressive cross-modal adaptation. To address these limitations, we propose GUIDE (Geometric Unrolling Inside MLLM Early-layers), a progressive framework for integrating geometric priors. GUIDE samples and aligns multi-level features from the geometric encoder and sequentially injects them into the early layers of the MLLM. This design enables the MLLM to continuously access multi-granularity geometric cues and progressively integrate them with visual representations during layer-wise semantic abstraction. GUIDE further introduces a dual context-aware gating mechanism at both the token and layer levels, selectively regulating geometric information to reduce redundant injection and interference with pretrained representations. Extensive experiments on spatial reasoning and 3D scene understanding benchmarks, including VSI-Bench, ScanRefer, and Scan2Cap, validate the effectiveness and cross-task applicability of GUIDE. Our 5B and 9B models achieve average scores of 71.5 and 72.1 on VSI-Bench, respectively.

cs.CV↗

arXiv2026

New upper bound for the Ramsey number of odd cycles

The \emph{$k$-color Ramsey number} $R_k(C_{2\ell+1})$ is the least integer $n$ such that any $k$-edge-coloring of a complete graph $K_n$ has a monochromatic odd cycle $C_{2\ell+1}$. Axenovich, Cames van Batenburg, Janzer, Michel, and Rundström~(JCT-B, 2026) recently proved \[ R_k(C_{2\ell+1})\le (4\ell-2)^k k^{k/\ell}+1, \] and Miyazaki, Mulrenin, Pohoata, and Zheng further improved the factor $k^{k/\ell}$ to $(k!)^{1/\ell}$. As Jenssen and Skokan (AM, 2021) determined $R_k(C_{2\ell+1})$ for fixed $k$ and sufficiently large $\ell$, it becomes even more interesting to seek better bound for fixed $\ell$ and sufficiently large $k$. In this paper, we show \[ R_k(C_{2\ell+1}) \le \frac{2\ell}{2\ell-1}(2\ell-1)^k(k!)^{1/\ell} \exp\!\left(k^{1-1/\ell}+O_\ell\!\left(k^{1-2/\ell}+\log k\right)\right)+1 \] for every fixed $\ell\ge 2$ and sufficiently large $k$, which improves the bound of Miyazaki et al. by a factor $2^{k-o(k)}$, and the bound of Axenovich et al. by a factor $(2\e^{1/\ell})^{k-o(k)}$.

math.CO↗

arXiv2026

ConsiSpace: Learning Geometric Consistency Matters for Video Spatial Reasoning

Video spatial reasoning is essential for navigation-oriented perception and long-video question answering, where models must infer spatial relations across long horizons under changing viewpoints. However, existing multimodal large language models (MLLMs) remain largely semantic-centric, and often fail to reliably aggregate consistent spatial evidence from redundant video observations, leading to inefficient or unstable reasoning. To address these issues, we propose ConsiSpace, a geometry-consistency-aware framework for geometry-sensitive video spatial reasoning that turns spatial consistency into both an evidence organization principle and an explicit post-SFT learning signal. We build a geometry-consistent memory (GCM) including implicit evidence tokens and explicit geometric cues, and leverage efficient organization strategies to compactly preserve task-related spatial evidence. Furthermore, we utilize unified consistency self-supervised reinforcement learning (UC-SSRL) after supervised fine-tuning to improve cross-view stability, with answer-, metric-, and topology-consistency rewards. Extensive experiments on three spatial-reasoning benchmarks, VSI-Bench, OSI-Bench, and MMSI-Video-Bench, show consistent gains, improving the average score by 12.6 points over the strongest baselines.

cs.CV↗

arXiv2026

A note on tree-cycle Ramsey numbers

Let $R(T_n,C_m)$ denote the Ramsey number of a tree $T_n$ on $n$ vertices versus a cycle $C_m$ of length $m$. Burr, Erdős, Faudree, Rousseau, and Schelp (1982) asked for the least function $f(m)$ such that $R(T_n,C_m)=2n-1$ for every odd $m\ge 3$ whenever $n\ge f(m)$. They proved that $f(m)\le 756m^{10}$. This bound was later improved to $25m$ by Brennan (2016) and to $4m-8$ by Fan and Lin (2025). In this note, we show that $f(m)\le 2m-4$ by using a different method and conjecture that $f(m)=\lceil (2m-1)/3\rceil$.

math.CO↗

arXiv2026

Phase amplification microscopy with femtometer-level accuracy

We demonstrate a major breakthrough in laser interferometry and microscopy achieving femtometer-level measurement accuracy and beyond, termed Phase Amplification microscopy (Φ-Amp). By exploiting the native silicide substrate as a phase cavity, our phase-gain theory predicts that weak sub-atomic phase signals can be magnified over 1000-fold, thus bypassing the shot-noise limit. We experimentally achieved a 158.2-fold phase gain for graphene in ambient air, corresponding ~ 730 femtometer accuracy. To fully unleash the potential of Φ-Amp for atomic fabrication and quantum measurement, we quantified interlayer spacing differences between AB-stacked and 30-degree-twisted bilayer graphene to be ~ 0.77 Angstroms and further detected atomic impurities and defects on large atomic structures. As the first wide-field metrology tool, we envision Φ-Amp may accelerate the scaling up of atomic quantum devices.

physics.optics↗

arXiv2026

Odd covers for complete graphs and complete 3-graphs

The Graham-Pollak theorem says that one needs at least $n - 1$ complete bipartite graphs to cover each edge of a complete graph $K_{n}$ on $n$ vertices exactly once. The odd cover problem is a parity analogue which seeks the minimum number of complete bipartite graphs, denoted by $b_2(n)$, such that each edge of $ K_n $ is covered an odd number of times. An odd cover of a complte 3-graph $K_n^{(3)}$ on $n$ vertices is a family of complete $3$-partite $3$-graphs such that every triple is covered an odd number of times. Let $b_3(n)$ be the minimum size of such a family. The values of $b_2(n)$ and $b_3(n)$ are determined for some $n$ in several previous works. In this paper, we first determine the value of $b_2(n)$ for all $n$, which confirms a conjecture due to Buchanan et al. (JGT, 2026), and then show $b_3(n+1)=b_2(n)$ by which the value of $b_3(n)$ is determined for all $n$, that resolves a question posed by Leader and Tan (EJC, 2026).

math.CO↗

arXiv2026

OpenGround: Planning-based Online Perception for Open-World 3D Visual Grounding

3D visual grounding aims to locate objects based on natural language descriptions in 3D scenes. Existing supervised methods are limited by generalization and recent zero-shot methods typically rely on a predefined Object Lookup Table (OLT) to query Visual Language Models (VLMs) for reasoning about object locations via a single step grounding, which limits the applications in scenarios with undefined targets and complex queries. To address these problems, we present OpenGround, a novel zero-shot framework for open-world 3D visual grounding that remains compatible with recent zero-shot methods. OpenGround integrates Task-Chain Planning to decompose a query into a plan of context-to-target sub-goals for progressive grounding, and Context-Guided Perception to perceive novel objects online under context guidance from the task chain. We also propose a new dataset named OpenTarget, which contains over 7000 object-description pairs to mimic open-world evaluation. Extensive experiments demonstrate that OpenGround achieves competitive performance on Nr3D, state-of-the-art on ScanRefer, and delivers a substantial 17.6\% improvement on OpenTarget. Project Page at https://why-102.github.io/openground.io/.

cs.CV↗