Regulating Anatomy-Aware Rewards via Trajectory-Integral Feedback for Volumetric Computed Tomography Analysis

Medical vision-language models (VLMs) have rapidly advanced as general-purpose multimodal assistants, yet their deployment in 3D Computed Tomography (CT) analysis remains constrained by a persistent mismatch between optimization objectives and clinical rigor. Current Reinforcement Learning (RL) paradigms still rely on lexical proxy signals that induce ``\textit{Evaluation Hallucinations}'', where models optimize linguistic fluency rather than factual clinical correctness, leading to diagnostically critical errors. To bridge this gap, we introduce the \textbf{Clinical Abnormality Benchmarking Substrate (CABS)}, a structured system that decomposes radiology reports into verifiable clinical semantic units. Using CABS, we identify a ``\textit{Mechanistic Divergence}'' in standard RL, where surface-similarity rewards drive policy gradients to bypass medical facts. We therefore propose \textbf{Trajectory-Integral Feedback GRPO (TIF-GRPO)}, a novel framework integrating control-theoretic principles into policy optimization. By formulating clinical reasoning as a pseudo-temporal trajectory for anomaly discovery, TIF-GRPO regulates anatomy-aware rewards via an integral feedback loop that penalizes persistent omissions as cumulative state errors and suppresses hallucinations as excessive control effort. Experiments on 3D CT benchmarks demonstrate that our approach significantly enhances abnormality detection and clinical faithfulness, establishing a new paradigm for fine-grained regulation in medical VLMs. Our project is available at \href{https://github.com/ZJU4HealthCare/TIF-GRPO}{GitHub}.

GEAR: Granularity-Adaptive Advantage Reweighting for LLM Agents via Self-Distillation

Reinforcement learning has become a widely used post-training approach for LLM agents, where training commonly relies on outcome-level rewards that provide only coarse supervision. While finer-grained credit assignment is promising for effective policy updates, obtaining reliable local credit and assigning it to the right parts of the long-horizon trajectory remains an open challenge. In this paper, we propose Granularity-adaptivE Advantage Reweighting (GEAR), an adaptive-granularity credit assignment framework that reshapes the trajectory-level GRPO advantage using token- and segment-level signals derived from self-distillation. GEAR compares an on-policy student with a ground-truth-conditioned teacher to obtain a reference-guided divergence signal for identifying adaptive segment boundaries and modulating local advantage weights. This divergence often spikes at the onset of a semantic deviation, while later tokens in the same autoregressive continuation may return to low divergence. GEAR therefore treats such spikes as anchors for adaptive credit regions: where the student remains aligned with the teacher, token-level resolution is preserved; where it departs, GEAR groups the corresponding continuation into an adaptive segment and uses the divergence at the departure point to modulate the segment' s advantage. Experiments across eight mathematical reasoning and agentic tool-use benchmarks with Qwen3 4B and 8B models show that GEAR consistently outperforms standard GRPO, self-distillation-only baselines, and token- or turn-level credit-assignment methods. The gains are especially strong on benchmarks with lower GRPO baseline accuracy, reaching up to around 20\% over GRPO, suggesting that the proposed adaptive reweighting scheme is especially useful in more challenging long-horizon settings.

What to Ignore, What to React: Visually Robust RL Fine-Tuning of VLA Models

Reinforcement learning (RL) fine-tuning has shown promise for Vision-Language-Action (VLA) models in robotic manipulation, but deployment-time visual shifts pose practical challenges. A key difficulty is that standard task rewards supervise task success, but offer limited guidance on whether a visual change is task-irrelevant or changes the behavior required for manipulation. We propose PAIR-VLA (Paired Action Invariance & Sensitivity for Visually Robust VLA), an RL fine-tuning framework to address this difficulty by adding two auxiliary objectives over paired visual variants during PPO optimization: an invariance term that reduces the discrepancy between action distributions for a task-preserving pair (e.g., different distractors), and a sensitivity objective that encourages separable action distributions for a task-altering pair (e.g., target object in a different pose). Together, these objectives turn visual variants from mere observation diversity into behavior-level guidance on policy responses during RL fine-tuning. We evaluate on ManiSkill3 across two representative VLA architectures, OpenVLA and $π_{0.5}$, under diverse out-of-distribution visual shifts including unseen distractors, texture changes, target object pose variation, viewpoint shifts, and lighting changes. Our method consistently improves over standard PPO, achieving average improvements of 16.62% on $π_{0.5}$ and 9.10% on OpenVLA. Notably, ablations further show generalization across visual shifts: invariance guidance learned from distractor and texture variants transfers to target-pose and lighting shifts, while adding sensitivity guidance on target-pose variants further improves robustness to nuisance shifts, highlighting the broader transferability of behavior-level RL guidance.

CT-FineBench: A Diagnostic Fidelity Benchmark for Fine-Grained Evaluation of CT Report Generation

The evaluation of generated reports remains a critical challenge in Computed Tomography (CT) report generation, due to the large volume of text, the diversity and complexity of findings, and the presence of fine-grained, disease-oriented attributes. Conventional evaluation metrics offer only coarse measures of lexical overlap or entity matching and fail to reflect the granular diagnostic accuracy required for clinical use. To address this gap, we propose CT-FineBench, a benchmark built from CT-RATE and Merlin to evaluate the fine-grained factual consistency of CT reports, constructed from CT-RATE and Merlin. Our benchmark is constructed through a meticulous, Question-Answering (QA) based process: first, we identify and structure key, finding-specific clinical attributes (like location, size, margin). Second, we systematically transform these attributes into a QA dataset, where questions probe for specific clinical details grounded in gold-standard reports. The evaluation protocol for CT-FineBench involves using this QA dataset to query a machine-generated report and scoring the correctness of the answers. This allows for a comprehensive, interpretable, and clinically-relevant assessment, moving beyond superficial lexical overlap to pinpoint specific clinical errors. Experiments show that CT-FineBench correlates better with expert clinical assessment and is substantially more sensitive to fine-grained factual errors than prior metrics.

Dictionary-based Pathology Mining with Hard-instance-assisted Classifier Debiasing for Genetic Biomarker Prediction from WSIs

Prediction of genetic biomarkers, e.g., microsatellite instability in colorectal cancer is crucial for clinical decision making. But, two primary challenges hamper accurate prediction: (1) It is difficult to construct a pathology-aware representation involving the complex interconnections among pathological components. (2) WSIs contain a large proportion of areas unrelated to genetic biomarkers, which make the model easily overfit simple but irrelative instances. We hereby propose a Dictionary-based hierarchical pathology mining with hard-instance-assisted classifier Debiasing framework to address these challenges, dubbed as D2Bio. Our first module, dictionary-based hierarchical pathology mining, is able to mine diverse and very fine-grained pathological contextual interaction without the limit to the distances between patches. The second module, hard-instance-assisted classfier debiasing, learns a debiased classifier via focusing on hard but task-related features, without any additional annotations. Experimental results on five cohorts show the superiority of our method, with over 4% improvement in AUROC compared with the second best on the TCGA-CRC-MSI cohort. Our analysis further shows the clinical interpretability of D2Bio in genetic biomarker diagnosis and potential clinical utility in survival analysis. Code will be available at https://github.com/DeepMed-Lab-ECNU/D2Bio.

Rethinking the Efficiency and Effectiveness of Reinforcement Learning for Radiology Report Generation

Radiologists highly desire fully automated AI for radiology report generation (R2G), yet existing approaches fall short in clinical utility. Reinforcement learning (RL) holds potential to address these shortcomings, but its adoption in this task remains underexplored. In this paper, we revisit RL in terms of data efficiency and optimization effectiveness for R2G tasks. First, we explore the impact of data quantity and quality on the performance of RL in medical contexts, revealing that data quality plays a more critical role than quantity. To this end, we propose a diagnostic diversity-based data sampling strategy that enables comparable performance with fewer samples. Second, we observe that the majority of tokens in radiology reports are template-like and diagnostically uninformative, whereas the low frequency of clinically critical tokens heightens the risk of being overlooked during optimization. To tackle this, we introduce Diagnostic Token-weighted Policy Optimization (DiTPO), which directly optimizes for clinical accuracy by using a diagnostic F1 score as the reward signal. Unlike standard RL approaches that treat all tokens equally, DiTPO explicitly models the varying importance of different tokens through rule- or gradient-based mechanisms to prioritize clinically relevant content. Extensive experiments on the MIMIC-CXR, IU-Xray, and CheXpert Plus datasets demonstrate that our framework achieves state-of-the-art (SOTA) performance while requiring substantially fewer training samples in RL. Notably, on MIMIC-CXR, our framework attains an F1 score of 0.516 using only 20% of the RL training samples.

OmniCT: Towards a Unified Slice-Volume LVLM for Comprehensive CT Analysis

Computed Tomography (CT) is one of the most widely used and diagnostically information-dense imaging modalities, covering critical organs such as the heart, lungs, liver, and colon. Clinical interpretation relies on both slice-driven local features (e.g., sub-centimeter nodules, lesion boundaries) and volume-driven spatial representations (e.g., tumor infiltration, inter-organ anatomical relations). However, existing Large Vision-Language Models (LVLMs) remain fragmented in CT slice versus volumetric understanding: slice-driven LVLMs show strong generalization but lack cross-slice spatial consistency, while volume-driven LVLMs explicitly capture volumetric semantics but suffer from coarse granularity and poor compatibility with slice inputs. The absence of a unified modeling paradigm constitutes a major bottleneck for the clinical translation of medical LVLMs. We present OmniCT, a powerful unified slice-volume LVLM for CT scenarios, which makes three contributions: (i) Spatial Consistency Enhancement (SCE): volumetric slice composition combined with tri-axial positional embedding that introduces volumetric consistency, and an MoE hybrid projection enables efficient slice-volume adaptation; (ii) Organ-level Semantic Enhancement (OSE): segmentation and ROI localization explicitly align anatomical regions, emphasizing lesion- and organ-level semantics; (iii) MedEval-CT: the largest slice-volume CT dataset and hybrid benchmark integrates comprehensive metrics for unified evaluation. OmniCT consistently outperforms existing methods with a substantial margin across diverse clinical tasks and satisfies both micro-level detail sensitivity and macro-level spatial reasoning. More importantly, it establishes a new paradigm for cross-modal medical imaging understanding.

Non-Contrast CT Esophageal Varices Grading through Clinical Prior-Enhanced Multi-Organ Analysis

Esophageal varices (EV) represent a critical complication of portal hypertension, affecting approximately 60% of cirrhosis patients with a significant bleeding risk of ~30%. While traditionally diagnosed through invasive endoscopy, non-contrast computed tomography (NCCT) presents a potential non-invasive alternative that has yet to be fully utilized in clinical practice. We present Multi-Organ-COhesion Network++ (MOON++), a novel multimodal framework that enhances EV assessment through comprehensive analysis of NCCT scans. Inspired by clinical evidence correlating organ volumetric relationships with liver disease severity, MOON++ synthesizes imaging characteristics of the esophagus, liver, and spleen through multimodal learning. We evaluated our approach using 1,631 patients, those with endoscopically confirmed EV were classified into four severity grades. Validation in 239 patient cases and independent testing in 289 cases demonstrate superior performance compared to conventional single organ methods, achieving an AUC of 0.894 versus 0.803 for the severe grade EV classification (G3 versus <G3) and 0.921 versus 0.793 for the differentiation of moderate to severe grades (>=G2 versus <G2). We conducted a reader study involving experienced radiologists to further validate the performance of MOON++. To our knowledge, MOON++ represents the first comprehensive multi-organ NCCT analysis framework incorporating clinical knowledge priors for EV assessment, potentially offering a promising non-invasive diagnostic alternative.

Holdout-Loss-Based Data Selection for LLM Finetuning via In-Context Learning

Fine-tuning large pretrained language models is a common approach for aligning them with human preferences, but noisy or off-target examples can dilute supervision. While small, well-chosen datasets often match the performance of much larger ones, systematic and efficient ways to identify high-value training data remain underexplored. Many current methods rely on heuristics or expensive retraining. We present a theoretically grounded, resource-efficient framework for data selection and reweighting. At its core is an In-Context Approximation (ICA) that estimates the holdout loss a model would incur after training on a candidate example by conditioning on a small, curated holdout set in context. ICA requires no reference model and no additional finetuning. Under a local linearization, ICA is equivalent to a first-order update toward the holdout optimum, motivating its use as a proxy for data value. We derive per-example weights from ICA scores, dynamically reweighting gradient updates as model parameters evolve. Across SFT, DPO, and SimPO, and over diverse backbones and datasets, ICA-based reweighting consistently improves model alignment with minimal overhead. We analyze sensitivity to score update frequency and the choice of $k$ holdout examples for in-context demonstrations, and note limitations for rapidly drifting on-policy updates, highlighting directions for future work. Code and prompts will be released.

HeurAgenix: Leveraging LLMs for Solving Complex Combinatorial Optimization Challenges

Heuristic algorithms play a vital role in solving combinatorial optimization (CO) problems, yet traditional designs depend heavily on manual expertise and struggle to generalize across diverse instances. We introduce \textbf{HeurAgenix}, a two-stage hyper-heuristic framework powered by large language models (LLMs) that first evolves heuristics and then selects among them automatically. In the heuristic evolution phase, HeurAgenix leverages an LLM to compare seed heuristic solutions with higher-quality solutions and extract reusable evolution strategies. During problem solving, it dynamically picks the most promising heuristic for each problem state, guided by the LLM's perception ability. For flexibility, this selector can be either a state-of-the-art LLM or a fine-tuned lightweight model with lower inference cost. To mitigate the scarcity of reliable supervision caused by CO complexity, we fine-tune the lightweight heuristic selector with a dual-reward mechanism that jointly exploits singals from selection preferences and state perception, enabling robust selection under noisy annotations. Extensive experiments on canonical benchmarks show that HeurAgenix not only outperforms existing LLM-based hyper-heuristics but also matches or exceeds specialized solvers. Code is available at https://github.com/microsoft/HeurAgenix.