arXiv subjects

Ziheng Ren

Publications and source records attributed to Ziheng Ren.

3 recordsLinked to original sources

arXiv2026

Semantic-Spatial Agreement Verification for Mitigating Object Hallucination in Multimodal Large Language Models

Multimodal large language models generate natural-language responses from visual inputs, yet may mention objects absent from an image. In medication assistance, accessible perception, and environmental decision-making, such hallucinations can create real-world safety risks. We propose Semantic-Spatial Agreement Verification (SSAV), a training-free method for verifying object claims. A visually grounded claim should remain stable across semantically equivalent queries and repeatedly localize to the same image region. SSAV aggregates multiple prompts to estimate semantic support and reduce sensitivity to query wording. Query-Induced Regional Verification (QIRV) combines cross-query region persistence, spatial overlap, and relative candidate dominance to identify isolated high responses and dispersed localizations. A geometric mean fuses semantic and spatial evidence, lowering the verification score when either branch lacks support. Experiments on three base models and multiple evaluation protocols show that SSAV effectively mitigates object hallucination. On LLaVA-1.5-7B, accuracy averaged across COCO, A-OKVQA, and GQA improves by 1.81 and 3.17 percentage points under POPE Popular and Adversarial, respectively, while CHAIRs decreases from 49.40% to 32.80%. These results show that cross-query semantic stability and regional consistency provide interpretable external visual evidence for object claims.

cs.CV↗

arXiv2026

FedSDR: Federated Self-Distillation with Rectification

Federated fine-tuning of Large Language Models faces severe statistical heterogeneity. However, existing model-level defenses often overlook the root cause: intrinsic data distribution mismatches. In this work, we first establish Federated Self-Distillation (FedSD) as a fundamental and potent strategy. By projecting client representations into a smoothed ``model-understanding space,'' FedSD alone serves as a universal booster, demonstrating superior performance over conventional algorithms. Despite its success, we identify a subtle trade-off termed the Rewrite Paradox -- unconstrained self-distillation can inadvertently increase hallucinations and redundancy. To refine this paradigm, we further propose FedSDR (Federated Self-Distillation with Rectification), the ultimate reinforced framework. It augments FedSD with a dual-stream mechanism: a local LoRA-S (Smoothing) branch to implicitly absorb heterogeneity via distilled data, and a parallel global LoRA-R (Rectification) branch anchored to raw data to enforce factual correctness. By selectively aggregating only LoRA-R, FedSDR yields a globally aligned and faithful model. Extensive experiments verify its superior performance.

cs.LG↗

arXiv2024

Ultrafast control of braiding topology in non-Hermitian metasurfaces

The mathematical theory of braids, influential across scientific disciplines, has emerged as a compelling strategy for light manipulation. Existing approaches to creating braids in photonics, whether in momentum-space bandstructures or real-space fields, often face limitations associated with static nature of devices and lack of tunability. Here, we experimentally demonstrate ultrafast control of eigen-spectrum braids of Jones matrices within mere picoseconds, in reconfigurable non-Hermitian metasurfaces. The Jones matrices of the metasurface exhibit a complex eigen-spectrum that braids in the three-dimensional eigenvalue-frequency space, thereby creating arbitrary elements within the two-string braid group, B2. By exciting the photoconductive semiconductor terahertz metasurface with a femtosecond infrared pulse, we achieve ultrafast switching of the braids, transitioning from the Solomon link to either the Trefoil knot or Hopf link. Our approach serves as a pivotal tool for elucidating non-trivial topology of braids and studying ultrafast topological optoelectronics.

physics.optics↗