Researcher profile

Tong Zheng

Tong Zheng contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate

Computation and Language Computer Vision Machine Learning Artificial Intelligence eess.IV Emerging Technologies physics.optics

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author

11works

0followers

7topics

4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification

Aligning Multimodal Large Language Models (MLLMs) requires reliable reward models, yet existing single-step evaluators can suffer from lazy judging, exploiting language priors over fine-grained visual verification. While rubric-based evaluation mitigates these biases in text-only settings, extending it to multimodal tasks is bottlenecked by the complexity of visual reasoning. The critical differences between responses often depend on instance-specific visual details. Robust evaluation requires dynamically synthesizing rubrics that isolate spatial and factual discrepancies. To address this, we introduce $\textbf{DeltaRubric}$, an approach that reformulates multimodal preference evaluation as a plan-and-execute process within a single MLLM. DeltaRubric operates in two steps: acting first as a $\textit{Disagreement Planner}$, the model generates a neutral, instance-specific verification checklist. Transitioning into a $\textit{Checklist Verifier}$, it executes these self-generated checks against the image and question to produce the final grounded judgment. We formulate DeltaRubric as a multi-role reinforcement learning problem, jointly optimizing planning and verification capabilities. Validated on Qwen3-VL 4B and 8B Instruct models, DeltaRubric achieves solid empirical gains. For instance, On VL-RewardBench, it improves base model overall accuracy by $\textbf{+22.6}$ (4B) and $\textbf{+18.8}$ (8B) points, largely outperforming standard no-rubric baselines. The results demonstrate that decomposing evaluation into structured, verifiable steps leads to more reliable and generalizable multimodal reward modeling.

preprint2026arXiv

Dual-Uncertainty Guided Policy Learning for Multimodal Reasoning

Reinforcement learning with verifiable rewards (RLVR) has advanced reasoning capabilities in multimodal large language models. However, existing methods typically treat visual inputs as deterministic, overlooking the perceptual ambiguity inherent to the visual modality. Consequently, they fail to distinguish whether a model's uncertainty stems from complex reasoning or ambiguous perception, preventing the targeted allocation of exploration or learning signals. To address this gap, we introduce DUPL, a dual-uncertainty guided policy learning approach for multimodal RLVR that quantifies and leverages both perceptual uncertainty (via symmetric KL divergence) and output uncertainty (via policy entropy) to guide policy updates. By establishing an uncertainty-driven feedback loop and employing a dynamic branch prioritization mechanism, DUPL recalibrates the policy advantage to focus learning on states with high perceptual or decisional ambiguity, enabling effective targeted exploration beyond passive data augmentation. Implemented on top of GRPO and evaluated on six multimodal mathematical and general-domain reasoning benchmarks, DUPL improves Qwen2.5-VL 3B and 7B models, achieving accuracy gains of up to 11.2% on visual math tasks and up to 7.1% on general-domain reasoning tasks, while consistently outperforming GRPO. These results demonstrate that dual-uncertainty guided policy learning is an effective and generalizable approach for multimodal RLVR.

preprint2026arXiv

G-Zero: Self-Play for Open-Ended Generation from Zero Data

Self-evolving LLMs excel in verifiable domains but struggle in open-ended tasks, where reliance on proxy LLM judges introduces capability bottlenecks and reward hacking. To overcome this, we introduce G-Zero, a verifier-free, co-evolutionary framework for autonomous self-improvement. Our core innovation is Hint-$δ$, an intrinsic reward that quantifies the predictive shift between a Generator model's unassisted response and its response conditioned on a self-generated hint. Using this signal, a Proposer model is trained via GRPO to continuously target the Generator's blind spots by synthesizing challenging queries and informative hints. The Generator is concurrently optimized via DPO to internalize these hint-guided improvements. Theoretically, we prove a best-iterate suboptimality guarantee for an idealized standard-DPO version of G-Zero, provided that the Proposer induces sufficient exploration coverage and the data filteration keeps pseudo-label score noise low. By deriving supervision entirely from internal distributional dynamics, G-Zero bypasses the capability ceilings of external judges, providing a scalable, robust pathway for continuous LLM self-evolution across unverifiable domains.

preprint2026arXiv

LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling

Test-time scaling (TTS) has become an effective approach for improving large language model performance by allocating additional computation during inference. However, existing TTS strategies are largely hand-crafted: researchers manually design reasoning patterns and tune heuristics by intuition, leaving much of the computation-allocation space unexplored. We propose an environment-driven framework, AutoTTS, that changes what researchers design: from individual TTS heuristics to environments where TTS strategies can be discovered automatically. The key to AutoTTS lies in environment construction: the discovery environment must make the control space tractable and provide cheap, frequent feedback for TTS search. As a concrete instantiation, we formulate width--depth TTS as controller synthesis over pre-collected reasoning trajectories and probe signals, where controllers decide when to branch, continue, probe, prune, or stop and can be evaluated cheaply without repeated LLM calls. We further introduce beta parameterization to make the search tractable and fine-grained execution trace feedback to improve discovery efficiency by helping the agent diagnose why a TTS program fails. Experiments on mathematical reasoning benchmarks show that the discovered strategies improve the overall accuracy--cost tradeoff over strong manually designed baselines. The discovered strategies generalize to held-out benchmarks and model scales, while the entire discovery costs only $39.9 and 160 minutes. Our data, and code will be open-source at https://github.com/zhengkid/AutoTTS.

preprint2026arXiv

Optical Quasi-symmetry Groups for Meron Lattices

We introduce quasi-symmetry groups in optics emerging from the commutation between mirror operation and the spin-orbit interaction (SOI) of light. Contrary to the principle of symmetry inheritance in free-space optics, where the symmetry of any structured field is strictly constrained by that of its source, we show that strong SOI enables quasi-symmetry-protected formation of meron lattices even when the underlying optical sources violate the nominal rotational symmetry. By analyzing the Hermiticity of the electric-dipole radiation amplitude in a circular polarization basis, we derive an effective mirror operator acting only on a subset of C3 polarized dipole emitters, forming a quasi-symmetry group that commutes with SOI. This quasi-symmetry guarantees exact C3 merons and gives rise to a robust polarization zone within which continuously varying input polarizations generate identical topological textures. Our work establishes quasi-symmetry as a new fundamental principle in optical physics and opens pathways to engineered topological structures of light beyond conventional symmetry constraints.

preprint2026arXiv

Reinforcing Multimodal Reasoning Against Visual Degradation

Reinforcement Learning has significantly advanced the reasoning capabilities of Multimodal Large Language Models (MLLMs), yet the resulting policies remain brittle against real-world visual degradations such as blur, compression artifacts, and low-resolution scans. Prior robustness techniques from vision and deep RL rely on static data augmentation or value-based regularization, neither of which transfers cleanly to critic-free RL fine-tuning of autoregressive MLLMs. Reinforcing reasoning against such corruptions is non-trivial: naively injecting degraded views during rollout induces reward poisoning, where perceptual occlusions trigger hallucinated trajectories and destabilize optimization. We propose ROMA, an RL fine-tuning framework that modifies the optimization dynamics to reinforce reasoning against visual degradation while preserving clean-input performance. A dual-forward-pass strategy uses teacher forcing to evaluate corrupted views against clean-image trajectories, avoiding new rollouts on degraded inputs. For distributional consistency, we apply a token-level surrogate KL penalty against the worst-case augmentation; to prevent policy collapse under regularization, an auxiliary policy gradient loss anchored to clean-image advantages preserves a reliable reward signal; and to avoid systematically incorrect invariance, correctness-conditioned regularization restricts enforcement to successful trajectories. On Qwen3-VL 4B/8B across seven multimodal reasoning benchmarks, our method improves robustness by +2.4% on seen and +2.3% on unseen corruptions over GRPO while matching clean accuracy.

preprint2026arXiv

RelayLLM: Efficient Reasoning via Collaborative Decoding

Large Language Models (LLMs) for complex reasoning is often hindered by high computational costs and latency, while resource-efficient Small Language Models (SLMs) typically lack the necessary reasoning capacity. Existing collaborative approaches, such as cascading or routing, operate at a coarse granularity by offloading entire queries to LLMs, resulting in significant computational waste when the SLM is capable of handling the majority of reasoning steps. To address this, we propose RelayLLM, a novel framework for efficient reasoning via token-level collaborative decoding. Unlike routers, RelayLLM empowers the SLM to act as an active controller that dynamically invokes the LLM only for critical tokens via a special command, effectively "relaying" the generation process. We introduce a two-stage training framework, including warm-up and Group Relative Policy Optimization (GRPO) to teach the model to balance independence with strategic help-seeking. Empirical results across six benchmarks demonstrate that RelayLLM achieves an average accuracy of 49.52%, effectively bridging the performance gap between the two models. Notably, this is achieved by invoking the LLM for only 1.07% of the total generated tokens, offering a 98.2% cost reduction compared to performance-matched random routers.

preprint2022arXiv

COVID-19 Infection Segmentation from Chest CT Images Based on Scale Uncertainty

This paper proposes a segmentation method of infection regions in the lung from CT volumes of COVID-19 patients. COVID-19 spread worldwide, causing many infected patients and deaths. CT image-based diagnosis of COVID-19 can provide quick and accurate diagnosis results. An automated segmentation method of infection regions in the lung provides a quantitative criterion for diagnosis. Previous methods employ whole 2D image or 3D volume-based processes. Infection regions have a considerable variation in their sizes. Such processes easily miss small infection regions. Patch-based process is effective for segmenting small targets. However, selecting the appropriate patch size is difficult in infection region segmentation. We utilize the scale uncertainty among various receptive field sizes of a segmentation FCN to obtain infection regions. The receptive field sizes can be defined as the patch size and the resolution of volumes where patches are clipped from. This paper proposes an infection segmentation network (ISNet) that performs patch-based segmentation and a scale uncertainty-aware prediction aggregation method that refines the segmentation result. We design ISNet to segment infection regions that have various intensity values. ISNet has multiple encoding paths to process patch volumes normalized by multiple intensity ranges. We collect prediction results generated by ISNets having various receptive field sizes. Scale uncertainty among the prediction results is extracted by the prediction aggregation method. We use an aggregation FCN to generate a refined segmentation result considering scale uncertainty among the predictions. In our experiments using 199 chest CT volumes of COVID-19 cases, the prediction aggregation method improved the dice similarity score from 47.6% to 62.1%.

preprint2022arXiv

Learning Multiscale Transformer Models for Sequence Generation

Multiscale feature hierarchies have been witnessed the success in the computer vision area. This further motivates researchers to design multiscale Transformer for natural language processing, mostly based on the self-attention mechanism. For example, restricting the receptive field across heads or extracting local fine-grained features via convolutions. However, most of existing works directly modeled local features but ignored the word-boundary information. This results in redundant and ambiguous attention distributions, which lacks of interpretability. In this work, we define those scales in different linguistic units, including sub-words, words and phrases. We built a multiscale Transformer model by establishing relationships among scales based on word-boundary information and phrase-level prior knowledge. The proposed \textbf{U}niversal \textbf{M}ulti\textbf{S}cale \textbf{T}ransformer, namely \textsc{Umst}, was evaluated on two sequence generation tasks. Notably, it yielded consistent performance gains over the strong baseline on several test sets without sacrificing the efficiency.

preprint2020arXiv

Multi-modality super-resolution loss for GAN-based super-resolution of clinical CT images using micro CT image database

This paper newly introduces multi-modality loss function for GAN-based super-resolution that can maintain image structure and intensity on unpaired training dataset of clinical CT and micro CT volumes. Precise non-invasive diagnosis of lung cancer mainly utilizes 3D multidetector computed-tomography (CT) data. On the other hand, we can take micro CT images of resected lung specimen in 50 micro meter or higher resolution. However, micro CT scanning cannot be applied to living human imaging. For obtaining highly detailed information such as cancer invasion area from pre-operative clinical CT volumes of lung cancer patients, super-resolution (SR) of clinical CT volumes to $μ$CT level might be one of substitutive solutions. While most SR methods require paired low- and high-resolution images for training, it is infeasible to obtain precisely paired clinical CT and micro CT volumes. We aim to propose unpaired SR approaches for clincial CT using micro CT images based on unpaired image translation methods such as CycleGAN or UNIT. Since clinical CT and micro CT are very different in structure and intensity, direct application of GAN-based unpaired image translation methods in super-resolution tends to generate arbitrary images. Aiming to solve this problem, we propose new loss function called multi-modality loss function to maintain the similarity of input images and corresponding output images in super-resolution task. Experimental results demonstrated that the newly proposed loss function made CycleGAN and UNIT to successfully perform SR of clinical CT images of lung cancer patients into micro CT level resolution, while original CycleGAN and UNIT failed in super-resolution.

preprint2020arXiv

Super-resolution of clinical CT volumes with modified CycleGAN using micro CT volumes

This paper presents a super-resolution (SR) method with unpaired training dataset of clinical CT and micro CT volumes. For obtaining very detailed information such as cancer invasion from pre-operative clinical CT volumes of lung cancer patients, SR of clinical CT volumes to $\m$}CT level is desired. While most SR methods require paired low- and high- resolution images for training, it is infeasible to obtain paired clinical CT and μCT volumes. We propose a SR approach based on CycleGAN, which could perform SR on clinical CT into $μ$CT level. We proposed new loss functions to keep cycle consistency, while training without paired volumes. Experimental results demonstrated that our proposed method successfully performed SR of clinical CT volume of lung cancer patients into $μ$CT level.

Institution

Affiliation not imported yet

This author record came from a source that does not expose affiliation metadata. Once the author claims the profile or we enrich the record from another provider, this section will link to the concrete institution.

Topic footprint