arXiv subjects

Jian Hu

Publications and source records attributed to Jian Hu.

At least 19 recordsLinked to original sources

arXiv2026

FIRM-Video: Check Before You Score for Reliable Text-to-Video Reward Modeling

Reliable reward models are essential for text-to-video evaluation and alignment. However, the trade-off between evaluation accuracy and inference efficiency places high demands on the quality of training supervision. Existing approaches often rely on holistic judges with fixed rubrics or open-ended reasoning, leading to incomplete inspection, unfaithful justification, and entangled attribution. We introduce FIRM-Video, a unified checklist-driven data construction framework based on a check-before-score principle: construct dimension-specific checklists, verify each criterion against temporal visual evidence, and aggregate only verified decisions. For Instruction Following, FIRM-Video decomposes prompts into weighted atomic requirements; for World Coherence, it constructs prompt-calibrated, target-specific checks grounded in visible entities and actions; and for Perceptual Quality, it applies a generic taxonomy of visual defects. The verified criteria and scores are further transformed into natural-language analyses for end-to-end reward modeling. Subsequently, we construct FIRM-Video-90K with 88,044 dimension-specific instances from 29,348 videos, and introduce FIRM-Video-Bench with 750 point-wise human annotations across 250 videos. The Qwen3-VL-based FIRM-Video-8B achieves the best overall MAE on FIRM-Video-Bench while consistently delivering the highest VBench Total, Quality, and Semantic Scores in Best-of-8 sampling across three video generators.

cs.CV↗

arXiv2026

EATR-Stereo: Embodiment-Aware Token Routing of Paired Stereo Evidence for Humanoid Vision-Language-Action Control

Long-horizon humanoid vision--language--action (VLA) control with head-mounted stereo cameras requires visual interfaces that can exploit complementary views while maintaining compatibility with pretrained representations. Existing interfaces often discard complementary stereo evidence or fuse additional observations without preserving the native primary-view pathway and adapting auxiliary information to robot embodiment. We present EATR-Stereo, an embodiment-aware token-routing framework that retains primary-view tokens and constructs primary-aligned Cross-View Auxiliary Tokens (CVATs) by querying the synchronized auxiliary-view token sequence. A body-segmented proprioceptive encoder further conditions token-wise auxiliary usage on robot configuration history, enabling selective incorporation of stereo evidence during action generation. The routed auxiliary stream augments the language and primary-visual context of a pretrained VLA while keeping its vision--language model frozen. On a 33-DoF physical humanoid with a 37-D proprioceptive state, we evaluate nine configurations in over-100-s search--approach--grasp--place--return tasks. EATR-Stereo achieves 60.0% full-task success, 100.0% grasp success, and 80.0% stage success. Under severe asymmetric occlusion, it improves recovery to 80% compared with 30% for CVAT alone. Ablation studies further show the importance of preserving primary tokens and combining cross-view auxiliary features with structured proprioceptive routing. These results demonstrate that selectively routed paired stereo evidence improves spatial grounding for reliable long-horizon humanoid VLA control.

cs.RO↗

arXiv2026

Molt: A Scalable PyTorch-Native Training Framework for Agentic Reinforcement Learning

Agentic reinforcement learning research is constant algorithm modification, new estimators, new pipeline stages, new rollout schemes, and in mainstream frameworks each change threads through layers of trainer, distributed backend, and rollout glue: the cost lands on the researcher at every iteration. Molt is a PyTorch-native training framework built to keep that cost small: a codebase compact and clean enough for a researcher to hold in their head, and for an AI coding assistant to read and reason about in its entirety, so the algorithm flow can be traced and changed end to end. The agent is an ordinary program, and one asynchronous loop trains multimodal and mixture-of-experts policies while never training on a token it did not generate, consistent in tokens, policy versions, and model semantics. Leanness does not cost performance: under a matched, fully asynchronous protocol, Molt is statistically comparable to a state-of-the-art Megatron-based stack. Molt is open source and provides recipes and containers at https://github.com/NVIDIA-NeMo/labs-molt.

cs.LG↗

arXiv2026

Beyond NL2Code: A Structured Survey of Multimodal Code Intelligence

While Large Language Models (LLMs) have substantially advanced text-to-code generation, many real programming tasks specify intent through visual artifacts such as screenshots, charts, and videos. These tasks require models to connect visual perception to executable programs, as correctness depends not only on syntax but also on layout, data semantics, and domain-specific constraints that apply after execution. This survey reviews Multimodal Code Intelligence, covering systems that generate, edit, refine, or reason with code under visually grounded inputs and outputs. We first formulate the field by the role that code plays in each task, distinguishing code as a rendered artifact, an editable structure, an intermediate reasoning trace, or an executable tool interface. Then we organize benchmarks and methods into four domains: Graphical User Interface, Scientific Visualization, Structured Graphics, and Frontier Tasks and Frameworks. This taxonomy connects artifact-generation problems to agentic and unified settings and allows us to compare how different tasks treat evidence of correctness. Across the literature, we argue that reliable evaluation requires evidence about semantics and interaction beyond visual fidelity. Looking ahead, future research may benefit from four verification-centered directions. Multi-signal validation can combine complementary evidence of correctness, multi-state verification can test behavior across execution trajectories, cross-task transfer testing can probe reusable visual-code skills, and verifiable agent traces can reveal whether agent actions are grounded in visual evidence. Together, these directions may move this field from single-output imitation toward evidence-grounded executable systems. An ongoing project and resources are available on \href{https://github.com/xjywhu/Awesome-Multimodal-LLM-for-Code}{GitHub}.

cs.CL↗

arXiv2026

Polar: Agentic RL on Any Harness at Scale

Reinforcement learning for language agents increasingly depends on custom harnesses that manage long-running context, multi-turn tool use and multi-agent orchestration. However, porting these harnesses into RL environment interfaces remains difficult and often loses important training signals. We bridge this gap with polar, a rollout framework for scalable asynchronous RL over arbitrary agent harnesses. Polar treats the agent harness as a black box: it proxies LLM API calls, records token-level model interactions, and reconstructs token-faithful trajectories for training. Each rollout node efficiently manages runtime prewarming, agent execution, trajectory reconstruction, and evaluation in parallel, exposing asynchronous service endpoints that can be consumed by independent trainers at scale. This decoupled design makes Polar agnostic to agent harnesses, training infrastructure, and RL algorithms while improving compute utilization for long-running agent workloads. We validate polar by training agents on software-engineering tasks with popular coding harnesses. Using simple GRPO, polar improves Qwen3.5-4B by 22.6, 4.8, 0.6 and 6.2 points on SWE-Bench Verified with the Codex, Claude Code, Qwen Code and Pi harnesses, respectively. We further demonstrate Polar for offline data generation over custom harnesses and ablate trajectory reconstruction strategies. Polar rewrites its preceding work, Prorl Agent, and has been registered as one of NeMo Gym environments.

cs.DC↗

arXiv2026

Reassessing the Statistical Necessity of Stellar Velocity Anisotropy in Strong-Lensing Cosmology with Lens-by-Lens Photometric Constraints

The stellar orbital anisotropy parameter ($\beta_{\rm ani}$) is a persistent systematic uncertainty in galaxy-scale strong gravitational lensing (SGL) cosmology. Typically fixed to isotropy or a local prior, it frequently degenerates with the lens density profile. We demonstrate this apparent redundancy largely arises from incomplete photometric constraints. We cross-matched 130 SGL systems with the Pantheon+ SN~Ia compilation, constructing a strictly matched sample of 107 SGL-SN pairs using a 5\% comoving-distance tolerance. Assuming a flat universe ($\Omega_k = 0$), the distance ratio is derived from apparent magnitude differences between paired SNe~Ia, eliminating $H_0$ and absolute magnitude dependence without fitting explicit dark-energy models. To break the kinematic degeneracy, we incorporate lens-by-lens luminosity density slopes ($\delta_i$) from high-resolution imaging. Adopting the quasi-model-independent P2 redshift-evolutionary framework ($\gamma(z) = \gamma_0 + \gamma_z z$), we find very strong statistical evidence for a free $\beta_{\rm ani}$. Fixing $\beta_{\rm ani}$ to isotropy ($\beta=0$) or a local prior ($\beta=0.18$) is strongly disfavored ($\Delta\bic = 14.2$ and $48.9$) and artificially inflates intrinsic scatter. A complementary P3 framework ($\gamma(z,\tilde{\Sigma}) = \gamma_0 + \gamma_z z + \gamma_s \log_{10}\tilde{\Sigma}$) confirms these penalties ($\Delta\bic = 13.5$ and $49.1$). Across all P2 variants, we consistently detect a negative redshift evolution of the density slope ($\gamma_z \approx -0.42$ to $-0.46$; ${\sim}1.5{-}2.0\sigma$), indicating ETG density profiles become shallower at higher redshifts. We conclude that when individual photometric constraints are incorporated, $\beta_{\rm ani}$ is statistically required as a free parameter to prevent severe dynamical modeling biases.

astro-ph.CO↗

arXiv2026

A model-independent test of the cosmic distance-duality relation using galaxy clusters and Type Ia supernovae matched pairs

The cosmic distance-duality relation (CDDR), expressed as $ D_L/D_A(1+z)^{-2}=1 $, is a fundamental relation in cosmology connecting luminosity distance ($ D_L $) and angular diameter distance ($ D_A $). Any departure from this relation would indicate new physics such as photon non-conservation, cosmic opacity, or non-metric gravity. We perform a stringent, model-independent test of the CDDR using a matched sample of 38 galaxy clusters from the Bonamente et al. compilation and Type Ia supernovae from the Pantheon+ sample. Employing the matched-pair technique, we simultaneously constrain the CDDR-violation parameter $ \eta $ and a possible redshift evolution of the SNe Ia absolute magnitude, parameterized as $ M_B(z)=M_0 + \varepsilon z $. We assess the robustness against matching tolerance and further supplement the analysis with DESI 2024 BAO measurements. Our results yield $ \eta = 0.050^{+0.348}_{-0.307} $ and $ \varepsilon = -0.184^{+0.724}_{-0.574} $ (68% CL), showing no statistically significant evidence for CDDR violation or SNe Ia evolution. The conclusions remain unchanged with stricter matching criteria and the inclusion of DESI BAO data. We also derive a cosmology-independent calibration of $ M_0 = -19.460^{+0.126}_{-0.124} $ mag. The standard cosmological model remains robust under this model-independent scrutiny.

astro-ph.CO↗

arXiv2026

ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents

Multi-turn LLM agents are increasingly important for solving complex, interactive tasks, and reinforcement learning (RL) is a key ingredient for improving their long-horizon behavior. However, RL training requires generating large numbers of sandboxed rollout trajectories, and existing infrastructures often couple rollout orchestration with the training loop, making systems hard to migrate and maintain. Under the rollout-as-a-service philosophy, we present ProRL Agent , a scalable infrastructure that serves the full agentic rollout lifecycle through an API service. ProRL Agent also provides standardized and extensible sandbox environments that support diverse agentic tasks in rootless HPC settings. We validate ProRL Agent through RL training on software engineering, math, STEM, and coding tasks. ProRL Agent is open-sourced and integrated as part of NVIDIA NeMo Gym.

cs.AI↗

arXiv2026

GraphThinker: Reinforcing Temporally Grounded Video Reasoning with Event Graph Thinking

Video reasoning requires a fine-grained understanding of the temporal dependencies and event-level relations between objects and events in videos. Current Multimodal Large Language Models (MLLMs) are prone to severe temporal hallucinations in video reasoning. An underlying cause of these hallucinations is weak visual-temporal grounding and the lack of explicit structure for modelling event relations. Models often rely on auxiliary text, such as dense captions, rather than explicitly anchoring their reasoning in actual visual evidence. However, these textual representations are inherently unstructured and fail to provide explicit causal constraints needed to guide the model's reasoning. In this work, we propose GraphThinker, a reinforcement finetuning method that constructs a structured event representation of a video and enforces visual grounding to jointly reduce reasoning hallucinations. Specifically, we employ an MLLM to construct an Event-based Video Scene Graph (EVSG) that captures both intra- and inter-event relations, guiding a structured video reasoning process. Moreover, we address the weak grounding issue by introducing a novel visual attention reward during reinforcement finetuning that encourages the model to actively attend to reliable visual cues. On the RexTime dataset, GraphThinker achieves an over 4% improvement in IoU=0.3 for moment localisation. On the VidHalluc dataset, GraphThinker achieves a 9.8% improvement in reducing temporal sequence hallucination and a 7.6% gain in Binary QA in reducing action hallucination, compared to the state-of-the-art methods.

cs.CV↗

arXiv2026

Golden Goose: A Simple Trick to Synthesize Unlimited RLVR Tasks from Unverifiable Internet Text

Reinforcement Learning with Verifiable Rewards (RLVR) has become a cornerstone for unlocking complex reasoning in Large Language Models (LLMs). Yet, scaling up RL is bottlenecked by limited existing verifiable data, where improvements increasingly saturate over prolonged training. To overcome this, we propose Golden Goose, a simple trick to synthesize unlimited RLVR tasks from unverifiable internet text by constructing a multiple-choice question-answering version of the fill-in-the-middle task. Given a source text, we prompt an LLM to identify and mask key reasoning steps, then generate a set of diverse, plausible distractors. This enables us to leverage reasoning-rich unverifiable corpora typically excluded from prior RLVR data construction (e.g., science textbooks) to synthesize GooseReason-0.7M, a large-scale RLVR dataset with over 0.7 million tasks spanning mathematics, programming, and general scientific domains. Empirically, GooseReason effectively revives models saturated on existing RLVR data, yielding robust, sustained gains under continuous RL and achieving new state-of-the-art results for 1.5B and 4B-Instruct models across 15 diverse benchmarks. Finally, we deploy Golden Goose in a real-world setting, synthesizing RLVR tasks from raw FineWeb scrapes for the cybersecurity domain, where no prior RLVR data exists. Training Qwen3-4B-Instruct on the resulting data GooseReason-Cyber sets a new state-of-the-art in cybersecurity, surpassing a 7B domain-specialized model with extensive domain-specific pre-training and post-training. This highlights the potential of automatically scaling up RLVR data by exploiting abundant, reasoning-rich, unverifiable internet text.

cs.AI↗

arXiv2025

BroRL: Scaling Reinforcement Learning via Broadened Exploration

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a key ingredient for unlocking complex reasoning capabilities in large language models. Recent work ProRL has shown promise in scaling RL by increasing the number of training steps. However, performance plateaus after thousands of steps, with clear diminishing returns from allocating more computation to additional training. In this work, we investigate a complementary paradigm for scaling RL, BroR-Lincreasing the number of rollouts per example to hundreds to exhaustively Broaden exploration, which yields continuous performance gains beyond the saturation point observed in ProRL when scaling the number of training steps. Our approach is motivated by a mass balance equation analysis allowing us to characterize the rate of change in probability mass for correct and incorrect tokens during the reinforcement process. We show that under a one-step RL assumption, sampled rollout tokens always contribute to correct-mass expansion, while unsampled tokens outside rollouts may lead to gains or losses depending on their distribution and the net reward balance. Importantly, as the number of rollouts per example N increases, the effect of unsampled terms diminishes, ensuring overall correct-mass expansion. To validate our theoretical analysis, we conduct simulations under more relaxed conditions and find that a sufficiently large rollout size N-corresponding to ample exploration-guarantees an increase in the probability mass of all correct tokens. Empirically, BroRL revives models saturated after 3K ProRL training steps and demonstrates robust, continuous improvement, achieving state-of-the-art results for the 1.5B model across diverse benchmarks.

cs.LG↗

arXiv2025

Taming Spontaneous Stop-and-Go Traffic Waves: A Bifurcation Perspective of A Dynamical Map

We consider a discrete-time dynamical system in a car-following context. The system was recently introduced to parsimoniously model human driving behavior based on utility maximization. The parameters of the model were calibrated using vehicle trajectory data from the Sugiyama experiment. It was shown that such a system can accurately reproduce the observed collective phenomena of a more elaborate experiment by Tadaki et al. Once the heterogeneity and noise are switched off, the model defines a map of the corresponding discrete-time dynamical system. We first perform a bifurcation analysis of the map by studying the stability of its limit solutions: a free-flow fixed point and a stop-and-go quasi-periodic orbit. When the vehicle density is varied, our model displays a bifurcation diagram qualitatively similar to those found in a class of optimal velocity models based on an ordinary differential equation approach, including regimes where one or both of the limit solutions are stable. In a 2D bifurcation diagram we further demonstrate that imposing a vehicle density-dependent speed advisory can dissipate the stop-and-go quasi-periodic orbit. This in turn lays the mathematical foundation for a simple, yet effective proposal [1] to tame stop-and-go waves, improving traffic flow and smoothness simultaneously via variable speed advisory.

eess.SY↗

arXiv2025

Part I: Tricks or Traps? A Deep Dive into RL for LLM Reasoning

Reinforcement learning for LLM reasoning has rapidly emerged as a prominent research area, marked by a significant surge in related studies on both algorithmic innovations and practical applications. Despite this progress, several critical challenges remain, including the absence of standardized guidelines for employing RL techniques and a fragmented understanding of their underlying mechanisms. Additionally, inconsistent experimental settings, variations in training data, and differences in model initialization have led to conflicting conclusions, obscuring the key characteristics of these techniques and creating confusion among practitioners when selecting appropriate techniques. This paper systematically reviews widely adopted RL techniques through rigorous reproductions and isolated evaluations within a unified open-source framework. We analyze the internal mechanisms, applicable scenarios, and core principles of each technique through fine-grained experiments, including datasets of varying difficulty, model sizes, and architectures. Based on these insights, we present clear guidelines for selecting RL techniques tailored to specific setups, and provide a reliable roadmap for practitioners navigating the RL for the LLM domain. Finally, we reveal that a minimalist combination of two techniques can unlock the learning capability of critic-free policies using vanilla PPO loss. The results demonstrate that our simple combination consistently improves performance, surpassing strategies like GRPO and DAPO.

cs.LG↗

arXiv2025

Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Temporal Grounding

Video Temporal Grounding (TG) aims to temporally locate video segments matching a natural language description (a query) in a long video. While Vision-Language Models (VLMs) are effective at holistic semantic matching, they often struggle with fine-grained temporal localisation. Recently, Group Relative Policy Optimisation (GRPO) reformulates the inference process as a reinforcement learning task, enabling fine-grained grounding and achieving strong in-domain performance. However, GRPO relies on labelled data, making it unsuitable in unlabelled domains. Moreover, because videos are large and expensive to store and process, performing full-scale adaptation introduces prohibitive latency and computational overhead, making it impractical for real-time deployment. To overcome both problems, we introduce a Data-Efficient Unlabelled Cross-domain Temporal Grounding method, from which a model is first trained on a labelled source domain, then adapted to a target domain using only a small number of unlabelled videos from the target domain. This approach eliminates the need for target annotation and keeps both computational and storage overhead low enough to run in real time. Specifically, we introduce. Uncertainty-quantified Rollout Policy Adaptation (URPA) for cross-domain knowledge transfer in learning video temporal grounding without target labels. URPA generates multiple candidate predictions using GRPO rollouts, averages them to form a pseudo label, and estimates confidence from the variance across these rollouts. This confidence then weights the training rewards, guiding the model to focus on reliable supervision. Experiments on three datasets across six cross-domain settings show that URPA generalises well using only a few unlabelled target videos. Codes will be released once published.

cs.CV↗

arXiv2025

Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training

Recent advancements in reasoning-focused language models such as OpenAI's O1 and DeepSeek-R1 have shown that scaling test-time computation-through chain-of-thought reasoning and iterative exploration-can yield substantial improvements on complex tasks like mathematics and code generation. These breakthroughs have been driven by large-scale reinforcement learning (RL), particularly when combined with verifiable reward signals that provide objective and grounded supervision. In this report, we investigate the effects of prolonged reinforcement learning on a small language model across a diverse set of reasoning domains. Our work identifies several key ingredients for effective training, including the use of verifiable reward tasks, enhancements to Group Relative Policy Optimization (GRPO), and practical techniques to improve training stability and generalization. We introduce controlled KL regularization, clipping ratio, and periodic reference policy resets as critical components for unlocking long-term performance gains. Our model achieves significant improvements over strong baselines, including +14.7% on math, +13.9% on coding, and +54.8% on logic puzzle tasks. To facilitate continued research, we release our model publicly.

cs.LG↗

arXiv2025

Beyond Charging Anxiety: An Explainable Approach to Understanding User Preferences of EV Charging Stations Using Review Data

Electric vehicles (EVs) charging infrastructure is directly related to the overall EV user experience and thus impacts the widespread adoption of EVs. Understanding key factors that affect EV users' charging experience is essential for building a robust and user-friendly EV charging infrastructure. This study leverages about $17,000$ charging station (CS) reviews on Google Maps to explore EV user preferences for charging stations, employing ChatGPT 4.0 for aspect-based sentiment analysis. We identify twelve key aspects influencing user satisfaction, ranging from accessibility and reliability to amenities and pricing. Two distinct preference models are developed: a micro-level model focused on individual user satisfaction and a macro-level model capturing collective sentiment towards specific charging stations. Both models utilize the LightGBM algorithm for user preference prediction, achieving strong performance compared to other machine learning approaches. To further elucidate the impact of each aspect on user ratings, we employ SHAP (SHapley Additive exPlanations), a game-theoretic approach for interpreting machine learning models. Our findings highlight the significant impact of positive sentiment towards "amenities and location", coupled with negative sentiment regarding "reliability and maintenance", on overall user satisfaction. These insights offer actionable guidance to charging station operators, policymakers, and EV manufacturers, empowering them to enhance user experience and foster wider EV adoption.

cs.HC↗

arXiv2025

ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models

Recent advances in reasoning-centric language models have highlighted reinforcement learning (RL) as a promising method for aligning models with verifiable rewards. However, it remains contentious whether RL truly expands a model's reasoning capabilities or merely amplifies high-reward outputs already latent in the base model's distribution, and whether continually scaling up RL compute reliably leads to improved reasoning performance. In this work, we challenge prevailing assumptions by demonstrating that prolonged RL (ProRL) training can uncover novel reasoning strategies that are inaccessible to base models, even under extensive sampling. We introduce ProRL, a novel training methodology that incorporates KL divergence control, reference policy resetting, and a diverse suite of tasks. Our empirical analysis reveals that RL-trained models consistently outperform base models across a wide range of pass@k evaluations, including scenarios where base models fail entirely regardless of the number of attempts. We further show that reasoning boundary improvements correlates strongly with task competence of base model and training duration, suggesting that RL can explore and populate new regions of solution space over time. These findings offer new insights into the conditions under which RL meaningfully expands reasoning boundaries in language models and establish a foundation for future work on long-horizon RL for reasoning. We release model weights to support further research: https://huggingface.co/nvidia/Nemotron-Research-Reasoning-Qwen-1.5B

cs.CL↗

arXiv2025

Agent RL Scaling Law: Agent RL with Spontaneous Code Execution for Mathematical Problem Solving

Large Language Models (LLMs) often struggle with mathematical reasoning tasks requiring precise, verifiable computation. While Reinforcement Learning (RL) from outcome-based rewards enhances text-based reasoning, understanding how agents autonomously learn to leverage external tools like code execution remains crucial. We investigate RL from outcome-based rewards for Tool-Integrated Reasoning, ZeroTIR, training base LLMs to spontaneously generate and execute Python code for mathematical problems without supervised tool-use examples. Our central contribution is we demonstrate that as RL training progresses, key metrics scale predictably. Specifically, we observe strong positive correlations where increased training steps lead to increases in the spontaneous code execution frequency, the average response length, and, critically, the final task accuracy. This suggests a quantifiable relationship between computational effort invested in training and the emergence of effective, tool-augmented reasoning strategies. We implement a robust framework featuring a decoupled code execution environment and validate our findings across standard RL algorithms and frameworks. Experiments show ZeroTIR significantly surpasses non-tool ZeroRL baselines on challenging math benchmarks. Our findings provide a foundational understanding of how autonomous tool use is acquired and scales within Agent RL, offering a reproducible benchmark for future studies. Code is released at \href{https://github.com/yyht/openrlhf_async_pipline}{https://github.com/yyht/openrlhf\_async\_pipline}.

cs.AI↗