Searcharxiv⌕ Search

arXiv subjects

Ming Kong

Publications and source records attributed to Ming Kong.

18 recordsLinked to original sources
arXiv2026

Don't Guess, Ask Wisely: Benchmarking and Training Proactive Elicitation in Open-Domain LLM Conversations

When user queries are incomplete, large language models should ask targeted questions before committing to an answer, a capability we call proactive elicitation, yet evaluation on real user-assistant queries across seven LLMs (8B to frontier) shows that fewer than 11% of underspecified queries trigger a follow-up question. To benchmark and train this capability, we introduce VagueBench, where each seed query is annotated with structured intent and detail units then degraded into intent-vague and fully-vague variants via controlled removal of linguistic cues, yielding known hidden information sets for evaluating both when and what to ask. We propose DialogGRPO, an offline reinforcement learning method that organizes multi-turn elicitation as a dialogue-state directed acyclic graph, where candidate questions at each node receive dense turn-level rewards combining immediate information gain with lookahead value through group-relative comparison, enabling fine-grained credit assignment without runtime simulation. DialogGRPO elicits missing information on over 88% of vague queries with substantial gains in both recovery efficiency and answer coverage, without over-asking on complete queries, outperforms prompting and test-time-scaling baselines, and generalizes zero-shot to MultiWOZ 2.4 and AmbigQA.

cs.CL↗
arXiv2026

D2R2: Discrete Diffusion with Regulation Reinforcement for Single-Cell Perturbation Prediction

Predicting single-cell transcriptomic responses to genetic perturbations is central to functional genomics and virtual-cell modeling. Existing approaches, however, typically predict an entire expression profile as a whole, leaving the order in which individual gene responses are generated unmodeled. To address this problem, we introduce \textbf{$D^{2}R^{2}$} (\textbf{D}iscrete \textbf{D}iffusion with \textbf{R}egulation \textbf{R}einforcement), which reformulates perturbation prediction as regulation-guided gene-wise progressive generation. A Masked Discrete Diffusion Model represents expression as ordinal tokens and reconstructs a fully masked profile step by step, allowing generated gene responses to condition those that remain masked. A Regulatory Policy Module initializes the generation policy from a gene regulatory network inferred from control cells and adapts it to the perturbation and current partially generated state. Then, group-relative policy optimization refines only the ordering policy using final perturbation-effect agreement as reward. Across Norman19 and VCC-H1, $D^{2}R^{2}$ achieves the best performance on all five metrics on Norman19 and remains competitive on H1. Controlled ablations holding the generator and generation budget fixed show that biological-prior ordering improves over random ordering and is more reliable than uncertainty-based heuristics, whereas reversing the biological-prior ordering degrades every metric. Biological analyses further show that the refined policy prioritizes regulatory genes early while promoting perturbation-specific transcription factors and responsive genes. These results establish gene generation order as an effective, controllable, and biologically interpretable dimension of single-cell perturbation prediction.

cs.AI↗
arXiv2026

REVEAL: Reverse Engineering Agentic Latents from Frontier-Model Traces

Frontier LLM agents perform complex interactive tasks, but their inference costs motivate transferring their experience to smaller open-source agents. Distillation and latent communication typically require target-model training or access to source-model internals. We study how textual frontier-model trajectories can provide reusable local controls for a frozen agent. Inspired by protocol reverse engineering, we introduce $\textit{Agentic Latents}$, comprising anchored states and dynamic profiles of feedback--response cycles elicited by replay in the target model. Our training-free framework, $\textbf{REVEAL}$ ($\textbf{REV}$erse $\textbf{E}$ngineering $\textbf{A}$gentic $\textbf{L}$atents), uses these profiles and trajectory outcomes to match and select local events from successful and failed rollouts of the same task and source model. Their response-transition contrasts yield a target-specific steering direction, applied before each response. Within-benchmark evaluations on Terminal-Bench 2.1 and Humanity's Last Exam with tools report higher accuracy than unmodified agents across three targets. Representation analyses and behavioral case studies support reusing frontier-model experience as local activation control.

cs.AI↗
arXiv2026

The Illusion of $\textit{What If}$: Evaluating the Breakdown of Counterfactual Reasoning in LLMs

Counterfactual reasoning requires models to reason beyond the observed world and explain how altered conditions propagate through downstream consequences. Existing benchmarks largely target bounded settings with fixed variables or single gold outcomes, overlooking open-domain scenarios requiring causal-process evaluation. To this end, we present $\textbf{WhatIfBench}$, a diagnostic benchmark for open-domain, open-form, long-horizon counterfactual causal reasoning, containing 220 what-if questions across STEM, HSS, and Hybrid scenarios. To evaluate free-form responses, we further propose $\textbf{PRISM}$, which first converts each natural-language explanation into a Response-Derived Semantic Causal Graph of events, states, and mechanisms. On top of this graph, PRISM then jointly applies a Process Metric assessing graph-level causal validity and a Rubric Metric assessing answer-level explanatory adequacy. Evaluating six frontier LLMs with this framework, we find that WhatIfBench remains far from saturated: even the strongest model reaches only a 64.62% final score. Further analysis reveals persistent causal gaps, premise drift, and topology fragmentation, suggesting that fluent counterfactual narratives often mask fragile causal processes. The benchmark, code, and evaluation scripts are available at $\href{https://github.com/zju-gt/WhatIfBench}{WhatIfBench}$.

cs.AI↗
arXiv2026

MetaRAG: Belief-Action Aligned Policy Optimization for Agentic RAG

Agentic retrieval-augmented generation (RAG) requires language models to decide when to continue searching and when to answer. Existing RL-based methods rely on external supervision and overlook the agent's internal belief about whether the current evidence is sufficient. To address this problem, we reformulate the search decision quality as belief-action alignment and propose MetaRAG, a belief-action aligned policy optimization framework for agentic RAG. MetaRAG uses Verify-first Action Generation to elicit an explicit verification process before each actual action, and Internal Belief Probing to estimate the policy model's own answerability belief from the same question-history context. Based on these, MetaRAG derives a consistency reward that is further gated by answer correctness, avoiding reinforcement of internally consistent but incorrect trajectories. The belief probe is used only during training and introduces no inference-time overhead. Experiments on seven public QA benchmarks show that MetaRAG consistently improves the accuracy-efficiency trade-off over strong RL-based agentic RAG baselines, with gains that transfer to deep research settings, different optimizers, and multiple model backbones.

cs.AI↗
arXiv2026

NGS-Marker: Robust Native Watermarking for 3D Gaussian Splatting

With the rapid development and adoption of 3D Gaussian Splatting (3DGS), the need for effective copyright protection has become increasingly critical. Existing watermarking techniques for 3DGS mainly focus on protecting rendered images via pre-trained decoders, leaving the underlying 3D Gaussian primitives vulnerable to misuse. In particular, they are ineffective against Partial Infringement, where an adversary extracts and reuses only a subset of Gaussians. In this paper, we propose NGS-Marker, a novel native watermarking framework for 3DGS. It integrates a jointly trained watermark injector and message decoder, and employs a gradientbased progressive injection strategy to ensure full-scene coverage. This enables robust ownership decoding from any local region. We further extend NGS-Marker with hybrid protection (combining native and indirect watermarks) and support for multimodal watermarking. Extensive experiments demonstrate that NGS-Marker effectively defends against partial infringement while offering practical flexibility for real-world deployment.

cs.CV↗
arXiv2026

Living-Harness Is an Interactive-Agent Evolver

Large language model (LLM) agents may recover from a failure within an episode or after a retry, yet the same execution failure can recur in later tasks because post-episode feedback rarely revises the persistent harness that guides future interactions. Static harnesses improve reliability through fixed tools, context, memory, and workflow structures, but remain unchanged after deployment. We propose $\textbf{Living-Harness}$, a self-evolving agent harness that converts each completed trajectory and its evaluator signals into posterior evidence for bounded harness updates. Guided by a domain-level $\textbf{Evolution-SOP}$ ($\textbf{S}$tandard $\textbf{O}$perating $\textbf{P}$rocedure), Living-Harness extracts an episode abstraction and structured update evidence, and writes two complementary forms of procedural knowledge: episodic memory that records trigger conditions, failure patterns, and recovery actions, and a state graph that records state nodes, repair edges, and transition rules. The updated harness state is retrieved to guide future interactions, while tools and base context remain frozen, allowing procedural repairs to accumulate across evolution cycles. On eight interactive environments derived from $τ^2$-Bench and MultiWOZ-2.4, Living-Harness improves average Pass@1 over the strongest interactive baseline by 10.07 and 9.91 percentage points, respectively, and supports retrieval-only reuse of the evolved harness state across model backbones. Our code will be made publicly available soon at https://github.com/anotherbricki/Living-Harness.

cs.MA↗
arXiv2026

CARE: Confidence-Aware Reasoning for Reliable Medical VQA

Reinforcement Fine-Tuning (RFT) has enabled medical Multimodal Large Language Models (MLLMs) to produce Chain-of-Thought (CoT) reasoning for visual question answering, yet these models suffer from $\textit{confidence miscalibration}$---a systematic gap between expressed certainty and actual diagnostic accuracy that undermines clinical trust. We propose $\textbf{CARE}$, a $\textbf{C}$onfidence-$\textbf{A}$ware medical $\textbf{RE}$asoning framework that jointly optimizes accuracy and calibration through a dual-stage pipeline. First, a scalable Medical-CoT synthesis provides structured cold-start data for Supervised Fine-Tuning. Second, Group Relative Policy Optimization (GRPO) with a novel $\textbf{Confidence-Aware Reward (CAR)}$ mechanism ties the model's confidence to diagnostic correctness within the reward signal. Across three Medical VQA benchmarks, $\textbf{CARE}$ achieves the highest diagnostic accuracy while obtaining the lowest Expected Calibration Error and Hallucination Rate, establishing a foundation for trustworthy clinical decision support. Our code is available at https://github.com/anotherbricki/CARE.

cs.CV↗
arXiv2026

CARL: Constraint-Aware Reinforcement Learning for Planning with LLMs

Despite their strong reasoning capabilities and extensive world knowledge, Large Language Models (LLMs) frequently generate plans that violate task constraints, undermining their reliability in real-world applications. This deficiency arises from a lack of systematic mechanisms to incorporate constraint information during the generation process. While existing approaches attempt to mitigate this by relying on external tools or task decomposition, they fail to enhance the model's intrinsic constraint awareness. To address this, we propose Constraint-Aware Reinforcement Learning (CARL), a novel RL framework designed to strengthen LLMs' intrinsic focus on constraints. CARL introduces a constraint-aware reward by comparing the model's output distributions under constrained and unconstrained inputs, encouraging constraint focus and penalizing neglect. Compatible with various RL frameworks and requiring no external solvers or top models, CARL enables scalable, end-to-end constraint-aware planning. Extensive experiments on BlocksWorld, TravelPlanner, and T-Eval demonstrate that CARL significantly outperforms standard Reinforcement Fine-Tuning (RFT) baselines and state-of-the-art reasoning models, exhibiting a markedly increased focus on constraints.

cs.AI↗
arXiv2026

STAPO: Selective Trajectory-Aware Policy Optimization for LLM Agent Training

Reinforcement Learning (RL) is the dominant paradigm for training Large Language Model (LLM) agents on long-horizon tasks. However, sparse and delayed rewards often lead to trajectory neglect, in which agents lose focus on the task goal and interaction history at intermediate steps. Prior work has explored step-level supervision using Shannon-entropy-based uncertainty signals, which conflate inherent state complexity with agent confidence and therefore provide unreliable estimates of decision reliability. To address this issue, we propose normalized entropy, which measures confidence deviations relative to an agent's average behavior under a given state, thereby strengthening the association between low-quality actions and trajectory neglect. Building on this insight, we introduce Selective Trajectory-Aware Policy Optimization (STAPO), a hierarchical group-based RL framework. STAPO leverages normalized entropy to locate outlier steps associated with trajectory neglect and optimizes them via a joint mechanism of trajectory-aware reward and trajectory-independent penalty, enhancing trajectory awareness while preserving training stability. Extensive experiments on ALFWorld, WebShop, and Search-Augmented QA demonstrate that STAPO achieves state-of-the-art performance while substantially alleviating trajectory neglect, validating its effectiveness and robustness for agentic tasks.

cs.AI↗
arXiv2026

Confidence Calibration for Multimodal LLMs: An Empirical Study through Medical VQA

Multimodal Large Language Models (MLLMs) show great potential in medical tasks, but their elicited confidence often misaligns with actual accuracy, potentially leading to misdiagnosis or overlooking correct advice. This study presents the first comprehensive analysis of the relationship between accuracy and confidence in medical MLLMs. It proposes a novel method that combines Multi-Strategy Fusion-Based Interrogation (MS-FBI) with auxiliary expert LLM assessment, aiming to improve confidence calibration in Medical Visual Question Answering (VQA). Experiments demonstrate that our method reduces the Expected Calibration Error (ECE) by an average of 40\% across three Medical VQA datasets, significantly enhancing MLLMs' reliability. The findings highlight the importance of domain-specific calibration for MLLMs in healthcare, offering a more trustworthy solution for AI-assisted diagnosis.

cs.CV↗
arXiv2026

AxiomOcean: Forecasting the Three-Dimensional Structure of the Upper Ocean

Short-term ocean forecast skill depends strongly on the three-dimensional ocean structure of the upper ocean, which governs stratification, subsurface heat storage, and the response of the ocean to atmospheric forcing. However, AI ocean forecasting models often fail to preserve this vertical structure, resulting in over-smoothed subsurface features and weak physical consistency under strong forcing. Here, we present AxiomOcean, a global AI ocean forecasting model that explicitly represents vertical hierarchy and cross-layer dependence within the water column. By combining a fully three-dimensional encoder-backbone-decoder architecture with surface atmospheric forcing, AxiomOcean jointly predicts upper-ocean temperature, salinity, and three-dimensional currents at global 1/12° resolution down to 643 m depth. In 10-day forecasts, AxiomOcean outperforms an advanced AI comparison model across variables and lead times, reducing day-1 RMSE by approximately 20 to 35% while maintaining higher anomaly correlation. The gain is not achieved through excessive smoothing: AxiomOcean better preserves eddy kinetic energy, temperature and salinity variance. Its advantage also extends through the water column and remains evident across the equatorial Pacific, Kuroshio Extension, and Southern Ocean, yielding a more realistic reconstruction of upper-ocean heat content. These results show that explicitly preserving upper-ocean three-dimensional structure can improve both forecast accuracy and physical fidelity in AI ocean prediction.

cs.LG↗
arXiv2026

Variation-aware Flexible 3D Gaussian Editing

Indirect editing methods for 3D Gaussian Splatting (3DGS) have recently witnessed significant advancements. These approaches operate by first applying edits in the rendered 2D space and subsequently projecting the modifications back into 3D. However, this paradigm inevitably introduces cross-view inconsistencies and constrains both the flexibility and efficiency of the editing process. To address these challenges, we present VF-Editor, which enables native editing of Gaussian primitives by predicting attribute variations in a feedforward manner. To accurately and efficiently estimate these variations, we design a novel variation predictor distilled from 2D editing knowledge. The predictor encodes the input to generate a variation field and employs two learnable, parallel decoding functions to iteratively infer attribute changes for each 3D Gaussian. Thanks to its unified design, VF-Editor can seamlessly distill editing knowledge from diverse 2D editors and strategies into a single predictor, allowing for flexible and effective knowledge transfer into the 3D domain. Extensive experiments on both public and private datasets reveal the inherent limitations of indirect editing pipelines and validate the effectiveness and flexibility of our approach.

cs.GR↗
arXiv2026

Linking Perception, Confidence and Accuracy in MLLMs

Recent advances in Multi-modal Large Language Models (MLLMs) have predominantly focused on enhancing visual perception to improve accuracy. However, a critical question remains unexplored: Do models know when they do not know? Through a probing experiment, we reveal a severe confidence miscalibration problem in MLLMs. To address this, we propose Confidence-Driven Reinforcement Learning (CDRL), which uses original-noise image pairs and a novel confidence-based reward to enhance perceptual sensitivity and robustly calibrate the model's confidence. Beyond training benefits, calibrated confidence enables more effective test-time scaling as a free lunch. We further propose Confidence-Aware Test-Time Scaling (CA-TTS), which dynamically coordinates Self-Consistency, Self-Reflection, and Visual Self-Check modules guided by confidence signals. An Expert Model acts in multiple roles (e.g., Planner, Critic, Voter) to schedule these modules and provide external verification. Our integrated framework establishes new state-of-the-art results with consistent 8.8% gains across four benchmarks. More ablation studies demonstrate the effectiveness of each module and scaling superiority.

cs.CV↗
arXiv2025

Style-Aligned Image Composition for Robust Detection of Abnormal Cells in Cytopathology

Challenges such as the lack of high-quality annotations, long-tailed data distributions, and inconsistent staining styles pose significant obstacles to training neural networks to detect abnormal cells in cytopathology robustly. This paper proposes a style-aligned image composition (SAIC) method that composes high-fidelity and style-preserved pathological images to enhance the effectiveness and robustness of detection models. Without additional training, SAIC first selects an appropriate candidate from the abnormal cell bank based on attribute guidance. Then, it employs a high-frequency feature reconstruction to achieve a style-aligned and high-fidelity composition of abnormal cells and pathological backgrounds. Finally, it introduces a large vision-language model to filter high-quality synthesis images. Experimental results demonstrate that incorporating SAIC-synthesized images effectively enhances the performance and robustness of abnormal cell detection for tail categories and styles, thereby improving overall detection performance. The comprehensive quality evaluation further confirms the generalizability and practicality of SAIC in clinical application scenarios. Our code will be released at https://github.com/Joey-Qi/SAIC.

cs.CV↗
arXiv2025

Distilling Multi-view Diffusion Models into 3D Generators

We introduce DD3G, a formulation that Distills a multi-view Diffusion model (MV-DM) into a 3D Generator using gaussian splatting. DD3G compresses and integrates extensive visual and spatial geometric knowledge from the MV-DM by simulating its ordinary differential equation (ODE) trajectory, ensuring the distilled generator generalizes better than those trained solely on 3D data. Unlike previous amortized optimization approaches, we align the MV-DM and 3D generator representation spaces to transfer the teacher's probabilistic flow to the student, thus avoiding inconsistencies in optimization objectives caused by probabilistic sampling. The introduction of probabilistic flow and the coupling of various attributes in 3D Gaussians introduce challenges in the generation process. To tackle this, we propose PEPD, a generator consisting of Pattern Extraction and Progressive Decoding phases, which enables efficient fusion of probabilistic flow and converts a single image into 3D Gaussians within 0.06 seconds. Furthermore, to reduce knowledge loss and overcome sparse-view supervision, we design a joint optimization objective that ensures the quality of generated samples through explicit supervision and implicit verification. Leveraging existing 2D generation models, we compile 120k high-quality RGBA images for distillation. Experiments on synthetic and public datasets demonstrate the effectiveness of our method. Our project is available at: https://qinbaigao.github.io/DD3G_project/

cs.CV↗
arXiv2024

Probablistic Restoration with Adaptive Noise Sampling for 3D Human Pose Estimation

The accuracy and robustness of 3D human pose estimation (HPE) are limited by 2D pose detection errors and 2D to 3D ill-posed challenges, which have drawn great attention to Multi-Hypothesis HPE research. Most existing MH-HPE methods are based on generative models, which are computationally expensive and difficult to train. In this study, we propose a Probabilistic Restoration 3D Human Pose Estimation framework (PRPose) that can be integrated with any lightweight single-hypothesis model. Specifically, PRPose employs a weakly supervised approach to fit the hidden probability distribution of the 2D-to-3D lifting process in the Single-Hypothesis HPE model and then reverse-map the distribution to the 2D pose input through an adaptive noise sampling strategy to generate reasonable multi-hypothesis samples effectively. Extensive experiments on 3D HPE benchmarks (Human3.6M and MPI-INF-3DHP) highlight the effectiveness and efficiency of PRPose. Code is available at: https://github.com/xzhouzeng/PRPose.

cs.CV↗
arXiv2020

Deep Sequential Feature Learning in Clinical Image Classification of Infectious Keratitis

Infectious keratitis is the most common entities of corneal diseases, in which pathogen grows in the cornea leading to inflammation and destruction of the corneal tissues. Infectious keratitis is a medical emergency, for which a rapid and accurate diagnosis is needed for speedy initiation of prompt and precise treatment to halt the disease progress and to limit the extent of corneal damage; otherwise it may develop sight-threatening and even eye-globe-threatening condition. In this paper, we propose a sequential-level deep learning model to effectively discriminate the distinction and subtlety of infectious corneal disease via the classification of clinical images. In this approach, we devise an appropriate mechanism to preserve the spatial structures of clinical images and disentangle the informative features for clinical image classification of infectious keratitis. In competition with 421 ophthalmologists, the performance of the proposed sequential-level deep model achieved 80.00% diagnostic accuracy, far better than the 49.27% diagnostic accuracy achieved by ophthalmologists over 120 test images.

eess.IV↗