Search arXiv⌕ Search

arXiv subjects

Qi Zhang

Publications and source records attributed to Qi Zhang.

At least 55 records · Page 3Linked to original sources
arXiv2026

WorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment Evolution

Scaling interactive environments is important for training long-horizon agents through reinforcement learning. These environments often require further testing and refinement after initial generation, and correcting defects in a task may also require updating its work context and verifier. However, existing construction pipelines offer limited support for such iterative refinement while keeping these components aligned. We introduce WorkRover, a framework for training long-horizon cowork agents through recursive environment evolution. Its harness-driven synthesis enables construction agents to jointly build and validate semantically aligned work contexts, executable tasks, and evidence-grounded verifiers, adapting their actions to workspace state and execution feedback rather than following a fixed construction workflow. To support environment evolution, we introduce Factorial Compatibility Replay (FCR), which uses rollout feedback to evaluate combinations of component revisions through dependency-closed replay and selects the smallest validated repair among tested candidates while preserving task requirements and evaluation constraints. To provide finer-grained training signals, we propose Verifier-to-Turn Credit Assignment (VTCA), which augments trajectory-level GRPO advantages with signals from deterministic workspace-progress checks and assigns a bounded share of positive-progress credit to prerequisite turns using recorded tool-output and artifact dependencies. Using WorkRover, we post-train Qwen3.5-35B-A3B to obtain WorkRover-35B-A3B and evaluate it on several cowork-agent benchmarks. Its reported scores exceed those of Qwen3.5-35B-A3B by 26.27 percentage points on JobBench, 29.32 percentage points on OfficeQA Pro, and 413.25 Elo points on GDPval v2. These results suggest that WorkRover offers a promising foundation for scalable training of long-horizon cowork agents.

cs.AI↗
arXiv2026

A Few More Tokens Ahead: Reinforcing Parallelism in Diffusion LLMs

Parallel generation is widely regarded as an inherent advantage of diffusion large language models. Yet our fixed-quota evaluation of four representative models reveals severe degradation across mathematics and code, with all scores below 15% at 16 tokens per step. Token-level probes identify local predictive potential near the observed prefix, especially in mathematical text. We propose RePara, which reinforces short consecutive token blocks through group-relative policy optimization and reduces rollout and log-probability recomputation costs. On LLaDA and SDAR, \ours improves GSM8K and MATH-500 accuracy over JustGRPO under reduced denoising budgets, with a 1.96x training-step speedup in our LLaDA/MATH setup.

cs.LG↗
arXiv2026

Eval4DiRec: A Unified and Systematic Evaluation Framework for Diffusion-based Recommender Systems

Leveraging the strong generative capabilities and stable training dynamics of diffusion models, diffusion-based recommender systems (RSs) have recently emerged as a novel recommendation paradigm, attracting increasing attention from both academia and industry. However, despite the rapid growth of diffusion-based RSs, a critical issue has emerged: the lack of a unified and systematic quantitative evaluation benchmark, which often results in irreproducible experimental results and unfair comparisons across studies due to inconsistent data processing, training configurations, inference procedures, and evaluation protocols. To address this challenge, we propose Eval4DiRec, the first unified and open-source evaluation framework specifically designed for diffusion-based RSs. Eval4DiRec supports 14 representative diffusion-based RS models across five different recommendation scenarios, providing consistent and reproducible experimental settings to systematically assess their performance. Built upon this framework, we conduct extensive empirical studies to benchmark these models under unified protocols. The results highlight the strong potential of diffusion models for recommendation while also revealing key factors and practical challenges that substantially affect their performance, thereby establishing a solid foundation to facilitate fair evaluation and guide future research in this promising field. Our code and data are available at: https://github.com/wangcong2001/Eval4DiRec.

cs.IR↗
arXiv2026

SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents

SWE-Bench Pro has emerged as a standard benchmark for evaluating software engineering agents on challenging repository-level tasks. However, our analysis work show that its evaluation is undermined by two sources of unreliability: reward hacking, enabled by leakage of gold solutions or hidden evaluation information, and task quality issues, including misleading problem statements and improperly scoped tests. These issues can inflate benchmark performance and obscure agents' true coding ability. We present SWE-Bench Pro Verified, a verified version of SWE-Bench Pro that addresses both problems. Our approach combines anti-hacking safeguards that eliminate major leakage channels without disrupting normal agent functionality, with task refinement that minimally corrects inconsistencies within flawed instances. Evaluations on SWE-Bench Pro Verified reveal that some models perform substantially worse than previously evaluated, suggesting that existing results on SWE-Bench Pro may overestimate real software engineering capability. SWE-Bench Pro Verified offers a more trustworthy benchmark for assessing software engineering agents.

cs.AI↗
arXiv2026

SOAR: Source-Aware Reinforcement Learning with Precise Credit Assignment for Search Agents

Reinforcement learning (RL) for search agents often suffers from evidence fabrication, a failure mode where agents bypass external information ($E_{ext}$) and fabricate an incorrect answer. We argue that this fabrication arises from an over-reliance on unreliable internal knowledge ($E_{int}$) as shortcuts due to the inherent opacity of reasoning trajectories. To address this, we propose SOAR (SOurce-Aware Reinforcement learning), a paradigm that transforms monolithic reasoning into source-explicit trajectories. SOAR (i) disentangles $E_{ext}$ and $E_{int}$ into parallel, observable streams via a Source-Aware Workflow, and (ii) incorporates Source-Consistent Reward to guide the policy evolution. By making information-source attribution explicit, SOAR enables precise credit assignment between different knowledge streams. Evaluations show that SOAR outperforms baselines on 3B and 7B backbones by 7.3% and 5.4%, while reducing evidence fabrication by 74.8%. These confirm that SOAR significantly improves performance on knowledge-intensive benchmarks while effectively mitigating evidence fabrication, fostering more faithful and grounded reasoning with enhanced robustness.

cs.CL↗
arXiv2026

ExplorationBench: Measuring AI Systems' Exploration in Verifiable Alien Worlds

Scientific discovery begins where known problems end. There, AI systems must engage in exploration: framing hypotheses, designing experiments, and iterating on the results. However, evaluating this ability is difficult: (1) how to verify whether a genuinely new hypothesis holds, and (2) how to determine whether a system has discovered it through exploration or merely recalled related knowledge from pre-training data. To this end, we introduce ExplorationBench, which turns the wicked problem of evaluating scientific exploration into a concrete and tractable framework built on verifiable Alien Worlds: their rules are executable, so every answer can be checked exactly, and they conflict with familiar knowledge, so recall alone cannot solve the tasks. The benchmark contains two sandboxes, AlienCode (31 discovery targets, 70 tasks) and AlienLogic (24 discovery targets, 70 tasks). Each sandbox provides a flawed manual, task-specific environmental feedback, and a dedicated tool-call schema. Systems use these resources to explore the sandbox, then solve held-out tasks. We evaluate 10 AI systems and find that the strongest systems can acquire and apply unfamiliar rules, while performance varies substantially across trajectories and continued exploration can stall or reverse earlier gains. ExplorationBench represents a step towards AI systems that can acquire and apply genuinely new knowledge through exploration in unknown environments.

cs.AI↗
arXiv2026

Gestalt: Large Multimodal Interplay Model

In this paper, we propose Gestalt, a new paradigm of large multimodal model built around multimodal interplay. Despite rapid advances, large multimodal models are reaching a bottleneck: existing approaches focus primarily on accommodating additional modalities while overlooking the distinct characteristics of each modality and the relations among them. Motivated by the multistage property of human multisensory perception, we propose a multimodal interplay pyramid that organizes multimodal modeling as a progression from modality-specific processing, through cross-modal alignment, to deeper multimodal integration. Guided by this pyramid, Gestalt adopts a unified discrete diffusion framework and an interplay-partitioned architecture, with learnable interplay tokens mediating cross-modal exchange and integration. The pyramid also structures its data organization and training strategy. Strong performance across image generation, multimodal understanding, and text-only evaluation shows that Gestalt significantly improves cross-modal integration while preserving modality-specific information, effectively harnessing the strengths of diffusion-based multimodal models and offering a promising path toward unified multimodal intelligence. Project page: https://GeWu-Lab.github.io/Gestalt.

cs.CV↗
arXiv2026

GLIO2: A GPU-Parallelized Tightly-Coupled LiDAR-Inertial-GNSS System for Robust and Real-Time Global Localization and Mapping

Globally consistent, real-time state estimation in large-scale, perceptually degraded environments is essential for autonomous vehicles and aerial robots, and requires fusing LiDAR, inertial, and GNSS measurements. Existing fusion methods, however, share a scan-to-map front-end with two failure modes. First, each scan is aligned to an incrementally built map that drifts under degeneracy, and once the estimate diverges the error is irrecoverable. Second, even without divergence, a registration biased by dynamic objects or wrong correspondences is propagated as a single pose constraint with an over-confident covariance, leaving its correspondences unavailable for GNSS to re-weight or relinearize. We propose GLIO2, a tightly-coupled LiDAR-Inertial-GNSS system whose GPU-parallel front-end jointly optimizes scan-to-multiscan LiDAR, IMU pre-integration, and raw GNSS measurements in a single sliding-window factor graph, sustaining real-time operation on edge hardware. A complementary offline back-end reuses the same cached factors to refine the entire trajectory in batch, completing the 30-min, 4.51-km UrbanNav Whampoa sequence in about 24 s. Across three public benchmarks (UrbanNav, MARS-LVIG, M3DGR) and self-collected UAV and vehicle data, GLIO2 attains the best overall accuracy among evaluated systems. On a 5.66-km bridge traversed at up to 96 km/h, where every competing baseline diverges under LiDAR degeneracy, it maintains 1.6 m horizontal accuracy. On an NVIDIA Jetson Orin NX, the full pipeline runs at about 25 Hz (39.60 ms per scan). The source code and datasets will be released.

cs.RO↗
arXiv2026

World Potential Model: Pretrained World Knowledge as Progress Potentials

Long-horizon language agents often receive supervision only from terminal task outcomes, leaving little signal for distinguishing productive intermediate behavior from stagnation or even regression. Rather than learning a separate value function or process reward model for every task, we ask whether pretrained models can recognize task progress from their existing world knowledge. We formalize this capability with a World Potential Model (WPM), a goal-conditioned evaluator of task-relative realized progress in agent contexts. In ALFWorld and ScienceWorld, off-the-shelf pretrained models substantially outperform chance at recovering realized-progress structure without task-specific evaluator fine-tuning. We further anchor these progress judgments to task-specific milestones to obtain scalar world potentials, whose temporal differences provide process-sensitive step-level credit for policy optimization. Under matched comparisons, WPM-guided optimization improves success over outcome-only GRPO across all evaluated configurations. Together, these results provide initial evidence that pretrained world knowledge can support reusable realized-progress evaluation and provide useful supervision for long-horizon agents.

cs.AI↗
arXiv2026

LiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving Markets

Evaluating agents by outcomes alone can obscure the capabilities that produce them. This problem is especially pronounced in evolving environments, where outcomes reflect a closed-loop interaction between agent behavior and changing external conditions. We introduce LiveMACEBench, a process-aware benchmark that uses live financial markets as a naturally evolving testbed for persistent LLM agents. Five frontier LLMs operate along continuous trajectories under matched Tool Use, Persistent Memory, Rule Following, and Multi-Agent Collaboration configurations. We evaluate them through both realized outcomes and mechanism-specific diagnostics derived from complete decision traces. Across 30 days of live evaluation, we find a pronounced outcome-capability gap: realized returns often diverge from capability-specific measurements, and similar outcomes can arise from markedly different patterns of mechanism use. Trace-level diagnostics further expose distinct bottlenecks across capabilities, demonstrating that mechanism access, effective mechanism use, and downstream performance are not interchangeable measures of agent capability. LiveMACEBench makes this distinction measurable, turning live markets from a performance leaderboard into a diagnostic environment for agent capability

cs.AI↗
arXiv2026

CATune: Structural Constraint-Aware Bayesian Optimization for DBMS Configuration Tuning

Modern DBMSs expose hundreds of configuration knobs, resulting in a high-dimensional and heterogeneous search space that makes automated tuning costly. Existing ML-based tuning systems typically treat the configuration domain as box-constrained and rely on workload feedback to implicitly capture inter-knob relationships. However, DBMS documentation specifies deterministic knob dependency constraints, particularly ordering constraints, that characterize structurally valid regions of the configuration space. We present CATune, a constraint-aware Bayesian optimization (BO) framework that models deterministic inter-knob ordering constraints as structural components of the search domain. Instead of learning feasibility boundaries through sampled violations, CATune performs optimization within a constraint-consistent subspace. We develop a topology-aware sampling strategy that respects dependency structure during exploration and avoids the inefficiencies of post-hoc constraint handling. To enable automated constraint discovery, we further design a precision-first extraction pipeline that combines LLM-based parsing with reliability safeguards to mitigate hallucinated dependencies. Experiments on PostgreSQL and MySQL using TPC-C and TPC-H workloads show that CATune substantially improves both sample efficiency and final tuning quality across surrogate models and BO frameworks. Under default ranges, CATune reaches the baseline optimum up to 12.5x faster and improves throughput by up to 63.37%. The improvements persist under knowledge-guided reduced ranges and alternative optimization implementations. These results demonstrate that explicitly modeling system-defined deterministic ordering constraints enhances optimization robustness and system stability.

cs.DB↗
arXiv2026

Affordance2Action: Task-Conditioned Scene-level Affordance Grounding for Real-Time Manipulation

Task-conditioned manipulation requires grounding instructions to task-relevant functional parts rather than object categories. This setting is scene-dependent and often one-to-many in cluttered scenes: the same object may afford different interactions across tasks, while a single task may correspond to either one functional region or multiple valid functional regions, depending on the scene layout. Existing affordance datasets and benchmarks remain misaligned with this setting, as they typically focus on grasping or object-level affordances, rely on synthetic scenes, or assume a single instruction-region correspondence. We present Affordance2Action (A2A), a benchmark-centered learning framework for scene-level, task-conditioned part affordance grounding. At its core is A2A-Bench, a manipulation-oriented benchmark that covers both single-region and multi-region instruction correspondences in everyday scenes, with the latter highlighting the ambiguity and diversity of affordance grounding in realistic multi-object environments. To construct it at scale, we build A2A-AffordGen, an agent-assisted annotation pipeline that combines language-model filtering, interactive part segmentation, instance-level mask-out refinement, task-reasoning instruction generation, and human verification. A2A-Bench's supervision further supports diverse downstream applications, with real-time affordance grounding and affordance-conditioned manipulation policies as two representative examples. Experiments show that A2A exposes substantial gaps in generic segmentation, VLM-based grounding, and affordance distillation baselines, while improving task-level localization and providing useful spatial priors for downstream manipulation.

cs.RO↗
arXiv2026

Finite-Momentum Antiferromagnetic Magnon Dynamics across the Morin Transition in Hematite

Hematite ($α$-Fe$_2$O$_3$) is a prototypical antiferromagnetic platform for high-speed spintronics and magnonics, yet its finite-momentum magnon dynamics and spin-reorientation mechanism remain incompletely understood. Here we use low-wavenumber magneto-Raman spectroscopy to resolve both $k=0$ and finite-$k$ sub-terahertz magnons in $α$-Fe$_2$O$_3$ across the Morin transition. We find that the limiting group velocity $v_0$ remains nearly field independent, whereas the finite-$k$ group velocity $v_{\mathrm g}$ is strongly modified near the Morin and spin-flop transitions. By combining parallel- and transverse-field measurements with spin-wave modeling, we further extract the temperature-dependent Dzyaloshinskii--Moriya field $H_{\mathrm D}$ and uniaxial anisotropy fields $H_{\mathrm{K1}}$ and $H_{\mathrm{K2}}$. We find that $H_{\mathrm D}$ is weakly temperature dependent at approximately 2.0--2.3~T, while $H_{\mathrm{K1}}$ decreases rapidly and $H_{\mathrm{K2}}$ changes only weakly. This contrast drives the sign reversal of the effective anisotropy field and quantitatively accounts for the Morin transition. Our results establish low-wavenumber Raman spectroscopy as a quantitative probe of finite-momentum antiferromagnetic magnons and provide key material parameters for hematite-based magnonics.

cond-mat.mtrl-sci↗
arXiv2026

Byte Language Models: Scaling, Emergent Abstractions, and Information Allocation

Tokenizer-free language models remove the inductive bias of fixed tokenizers by modeling text directly as bytes, but the resulting longer sequences substantially increase computation and eliminate explicit text abstractions. We ask whether this additional computation can be useful, and whether standard Transformers can learn the abstractions that tokenization provides. We study these questions on Transformers without specialized tokenization-related architectures. With token-superposition training and hash embeddings, byte Transformers consistently outperform subword Transformers as model size scales. We further find that byte Transformers build local text abstractions as external tokenizers: a set of segmentation-like positions are used to collect local context representations, and restricting up to $25\%$ of intermediate layers to these local representations preserves downstream performance. Finally, these learned structures induce highly non-uniform generation difficulty, with uncertainty concentrated near local structure boundaries; exploiting them for speculative decoding yields $3.4\times$ more accepted tokens than in subword Transformers.

cs.CL↗
arXiv2026

Compactly Supported Strong Ill-Posedness of the Two-Dimensional Euler Flow in $W^{3,1}(\mathbb R^2)$

We prove strong ill-posedness for the two-dimensional incompressible Euler equation in the endpoint velocity space $W^{3,1}_σ(\mathbb R^2)$. Given any compactly supported smooth divergence-free background and any $δ>0$, we construct a single compactly supported divergence-free datum within $δ$ of the background in $W^{3,1}$. The perturbation can be localized inside any prescribed nonempty open neighborhood of the background support while remaining disjoint from that support. The associated unique global finite-energy solution $u$ satisfies ${\rm curl} u\in C([0,\infty);W^{2,1}(\mathbb R^2))$, but \[ \operatorname*{ess\,sup}_{0 0. \] The proof combines localized hyperbolic insertion, the Bonami--Poornima nonmultiplier mechanism, and many-packet dilution. Moment cancellation and positive material separation control the nonlocal Biot--Savart interactions, allowing a nested construction inside one fixed compact region. Without the compact-support requirement on the limiting datum, a complementary far-field argument yields a dense $G_δ$ set of data whose solutions leave $W^{3,1}$ at every positive rational time and on a residual set of times, with unbounded essential supremum on every nonempty open time interval.

math.AP↗
arXiv2026

HyperLogic: A Hard, Forward-Authored Chinese Logical Reasoning Benchmark with Execution-Derived Answers

Existing logic benchmarks primarily measure models' ability to answer reasoning questions directly. Scalable benchmarks often generate text from formal structures, which makes answers easy to compute but fixes the formalization before the problem is written. Forward construction preserves the challenge of finding a faithful formalization, yet makes difficulty and answer reliability harder to control. We introduce HyperLogic, a forward-construction pipeline that separates problem authoring from answer generation. A multi-agent workflow hardens undergraduate-authored Chinese seeds without solving them; two agents from different model families independently translate each finished item into executable finite-domain models; their encodings and solver-derived answers undergo layered, agent-assisted adjudication under human-expert oversight. HyperLogic-Base contains 195 items and 922 sub-questions and separates seven frontier models by 33.0 percentage points in strict item accuracy (44.6-77.6%). HyperLogic-Hard contains 100 items with larger, coupled search spaces, on which no model exceeds 16% accuracy in direct answering. We also use Hard to evaluate agents' ability to formalize and solve problems with tools, comparing a code sandbox alone with one that includes our logic modeling library. The sandbox improves every model by 16.7-40.1 points; adding the library helps five models and hurts two. These results highlight the difficulty of faithful formalization even with tool access.

cs.CL↗
arXiv2026

MuseCritic: Learning Multi-Aspect Song Rewards through Natural-Language Aesthetic Critiques

Long-form song generation models continue to improve in duration, structural coherence, and acoustic complexity, increasing the need for reliable aesthetic rewards aligned with human preferences. However, reward models for complete songs remain limited, and existing evaluators typically predict scores in a single forward pass without readable explanations. To this end, we introduce MuseCritic, a semi-scalar reward model that generates a natural-language critique covering five aesthetic dimensions and uses it as an intermediate representation to predict continuous reward scores. MuseCritic follows a two-stage training pipeline: a teacher model first provides high-quality critiques for supervised fine-tuning, then the fine-tuned model generates its own critiques for reward learning, mitigating training-inference distribution shift. On an in-domain test set of 200 SongEval songs, MuseCritic reduces macro-averaged mean squared error from 0.2875 to 0.2316 and improves macro-averaged LCC, SRCC, and Kendall's tau to 0.9068, 0.8838, and 0.7178, respectively. On the out-of-domain Music Arena benchmark with 733 preference pairs, it achieves 71.35% accuracy and remains competitive with strong music-specific reward models. Using MuseCritic with GRPO also improves Muse-0.6B on all nine aesthetic metrics from SongEval and Audiobox Aesthetics. These results show that critique-conditioned reward modeling reduces scoring error and provides an effective optimization signal for song generation. The project repository is available at https://github.com/WuqnEl/MuseCritic.

cs.SD↗
arXiv2026

VisAudit: Evaluating Multimodal Agents for Visual Diagnosis and Repair

Multimodal agents are increasingly used for data visualization tasks but remain limited in autonomous review. Unlike humans, they may fail to recognize when a visualization is incorrect, determine what to change, repair it without disrupting correct content, and verify whether the intervention succeeded. Existing benchmarks largely evaluate predefined individual capabilities such as chart generation, instruction-guided editing, or defect detection, and therefore do not capture this gap in autonomous review. We introduce VisAudit, a benchmark for evaluating visualization diagnosis, repair, and verification. Given a rendered chart and configurable auxiliary evidence, including its source data table, intended text summary, and visualization code, an agent iteratively diagnoses potential defects, modifies and executes visualization code, inspects execution and visual feedback, and determines when no further intervention is needed. VisAudit defines three tracks spanning diagnosed repair, autonomous repair, and open-world verification, and contains 1,900 flawed instances across 21 chart types and 10 flaw categories, together with 300 initially correct charts. We construct the benchmark through controlled perturbations of validated source visualizations, with systematic verification and human-aligned quality control to ensure that injected defects are well-defined and recoverable from the available evidence. Experiments with leading multimodal models reveal a substantial gap from reliable autonomous review: the strongest evaluated model fully recovers only $47.4\%$ of flawed charts in the autonomous-repair setting.

cs.LG↗