Searcharxiv⌕ Search

arXiv subjects

Jiajun Zhang

Publications and source records attributed to Jiajun Zhang.

At least 37 records · Page 2Linked to original sources
arXiv2026

PhotoHOI: Synthesizing 3D Hand-Object Interactions from a Single RGB Photograph

Hand-object interaction (HOI) is a fundamental human behavior with broad applications in AR/VR, digital humans, and embodied interaction. Existing methods typically require predefined object geometry, object trajectories, or task-specific conditions, limiting their use with natural real-world inputs. To address this, we study a more practical problem of synthesizing 3D hand-object interaction sequences from a single RGB photograph and an open-vocabulary language instruction, and introduce PhotoHOI. PhotoHOI first uses a vision-language model to parse the input image and instruction into a structured task specification, including the interaction object, target region, and spatial relation. It then recovers a compact task-relevant 3D scene and plans a smooth collision-aware object trajectory based on the recovered object states, support relations, and surrounding scene geometry. To synthesize hand motion that generalizes to real-world photographs and unseen objects, it learns transferable task-conditioned contact and contact-conditioned grasp priors from large-scale affordance and HOI data. The grasp is further refined in a learned latent space, constraining the optimization to a plausible hand-pose manifold. Experiments on GRAB and H2O demonstrate improved contact quality and reduced penetration over representative baselines. Results on real-world photographs further demonstrate higher task success and scene consistency, together with generalization to unseen objects and open-vocabulary instructions.

cs.CV↗
arXiv2026

EviGraph: Evidence-Guided Autonomous Research Agents

Autonomous research agents can generate hypotheses, execute experiments, and draft manuscripts, yet their outputs often contain unsupported claims and inconsistencies between research questions, experiments, results, and conclusions. We argue that this problem is partly architectural: existing systems organize research as sequential pipelines but do not explicitly maintain or validate the evolving claim-evidence structure across stages. In this paper, we introduce EviGraph, an autonomous research framework that represents the research process as a typed evidence graph containing Problem, Gap, Hypothesis, Experiment, Finding, and Claim nodes. The graph serves as the operational state of the agent rather than a post-hoc record. EviGraph inspects evidence chains for missing dependencies, semantic misalignment, and result-claim inconsistencies, localizes the earliest weak node, and regenerates its affected downstream subgraph. Graph checkpointing prevents unsuccessful repairs from corrupting previously validated evidence. Manuscripts are generated only after every retained claim is grounded in a validated evidence chain. Experiments on ARC-Bench-ML and NanoResearch-20 show that EviGraph outperforms the compared end-to-end research-agent baselines in overall research performance, improves Claim Support Rate by 40.19% over the strongest baseline, and achieves 87.73% Experimental Data Consistency. These results demonstrate the value of explicit evidence-state maintenance for reliable autonomous research.

cs.AI↗
arXiv2026

Project2Task: Graph-Guided Project-Level Planning for Autonomous Research

Research agents can increasingly search literature, propose hypotheses, generate code, run experiments, and draft manuscripts from a single topic. However, a research project is not merely a larger task: it is a long-horizon agenda that must be advanced through multiple bounded tasks with distinct but related objectives, parallel alternatives, and dependency-aware sequences. Existing single-task systems often treat the project as one oversized task, produce a flat set of vague or overlapping tasks, or leave task boundaries and execution order to manual coordination. We introduce Project2Task, a graph-guided project-level planning layer for autonomous research. Given a project brief, it represents candidate contributions as innovation atoms and organizes them in a directed lineage graph. A lightweight Bernoulli block-model objective selects among horizontal, vertical, and hybrid portfolio decompositions. Project2Task then generates bounded tasks with explicit contribution ownership, repairs overlaps and missing execution fields, and emits dependency-aware task contracts that specify objectives, inputs, expected artifacts, evaluation requirements, boundary constraints, dependencies, and execution order. The contracts are independent of any particular downstream research executor and support integration of task outputs into a coherent project-level result. On a benchmark of ten project briefs yielding roughly 30 tasks, manuscript-based portfolio evaluation gives Project2Task an average quality score of 7.15, compared with 4.58 for the Brief Baseline and 5.31 for the Topic-only Setting. Integrating its contracts with AutoResearchClaw increases average downstream task accuracy from 0.536 to 0.759. These results demonstrate the value of explicit project-to-task planning for producing coherent, non-redundant, and executable research-task portfolios.

cs.AI↗
arXiv2026

Qwen-CUA: Native Computer Use for (almost) Everything

Native computer use offers a general interface for agents to operate almost any software available to people, but requires long-horizon state tracking, large-scale interactive experience, and learning from sparse yet verifiable outcomes. We introduce Qwen-CUA, a native computer-use agent with a 397B-A17B Qwen mixture-of-experts backbone. It observes only screenshots and acts through keyboard and mouse events, without DOM trees, accessibility metadata, or task-specific APIs. Its scaffold maintains up to 20 active screenshots and folds older visual history in fixed-size blocks to retain recent evidence while preserving reusable prompt prefixes. For training, we build a cloud rollout fleet with access to nearly 100,000 vCPUs and tens of thousands of concurrent environments, construct approximately 40,000 verifiable tasks, and collect personalized long-horizon workflows across everyday and professional software. We optimize complete trajectories with verifiable rewards and trajectory slicing, while iterative training runs refresh supervised data and recalibrate reinforcement-learning tasks. Across eight benchmarks, Qwen-CUA outperforms Qwen3.7 and remains competitive with leading proprietary systems, reaching 86.2 on OSWorld-Verified and 18.5/48.4 binary/partial completion on OSWorld 2.0. Scaling the same recipe to a model with over one trillion parameters yields Qwen-CUA-Max, improving these scores to 87.6 and 21.2/53.3. Qwen-CUA also reduces RedTeamCUA attack success from 36.6 to 16.4 relative to Qwen3.7. Efficiency analyses, a browser deployment, and Bash-augmented experiments further characterize practical behavior. These results establish native computer use as a broadly capable agent foundation and highlight scalable verifiable interaction and hybrid tool use as key directions.

cs.LG↗
arXiv2026

GigaSpeechBench: A Real-World Multilingual Speech-to-Text Benchmark

While modern ASR systems achieve low error rates on high-resource benchmarks, such performance often overestimates real-world robustness. Existing evaluations address challenges in isolation, lacking a unified benchmark for domain terminology, age variation, dialects, accents, and low-resource languages, particularly across the Middle East and Southeast Asia, representing over one billion under-evaluated speakers. To address this gap, we introduce GigaSpeechBench, a comprehensive multilingual and multidimensional in-the-wild ASR & AST benchmark comprising 680 hours of human-annotated speech. It features five modules: (1) 12 low-resource Middle Eastern and Southeast Asian languages, plus challenging Japanese and Korean; (2) 6 Chinese dialects; (3) 6 English accents; (4) dense terminology across 12 vertical domains for Chinese and English; and (5) older adult and child speech. We further provide human-annotated Chinese and English translations for 11 languages to support AST evaluation. Extensive evaluations of leading foundation models and commercial APIs reveal significant performance degradation in these challenging settings, exposing critical evaluation blind spots.

eess.AS↗
arXiv2026

Gravitational Wave Astronomy With TianQin

The opening of the gravitational wave window has significantly enhanced our capacity to explore the universe's most extreme and dynamic sector. In the mHz frequency range, a diverse range of compact objects, from the most massive black holes at the farthest reaches of the Universe to the lightest white dwarfs in our cosmic backyard, generate a complex and dynamic symphony of gravitational wave signals. Once recorded by gravitational wave detectors, these unique fingerprints have the potential to decipher the birth and growth of cosmic structures over a wide range of scales, from stellar binaries and stellar clusters to galaxies and large-scale structures. The TianQin space-borne gravitational wave mission is scheduled for launch in the 2030s, with an operational lifespan of five years. It will facilitate pivotal insights into the history of our universe. This document presents a concise overview of the detectable sources of TianQin, outlining their characteristics, the challenges they present, and the expected impact of the TianQin observatory on our understanding of them.

astro-ph.GA↗
arXiv2026

SWITCH: Benchmarking Modeling and Handling of Tangible Interfaces in Long-horizon Embodied Scenarios

Tangible control interfaces (TCIs), such as appliance panels, remotes, elevators, and embedded GUIs, are a fundamental component of everyday human-built environments. Interacting with these interfaces requires agents not only to ground language in visual observations,but also to execute actions, track temporally evolving state changes, and verify whether intended outcomes have been achieved. However, existing benchmarks predominantly evaluate open-loop perception or single-step action execution, failing to capture this continuous cycle of interaction, feedback, and correction. We introduce SWITCH, a benchmark for closed-loop interactive reasoning with TCIs in realistic egocentric environments1. SWITCH comprises 1,170 temporally interactive videos across diverse functional categories, providing structured annotations of instructions, actions, state transitions, outcomes, and recovery behaviors over time. To probe generative world modeling, SWITCH also evaluates video generation models on interaction-centered tasks using both LLM-as-judge and human evaluation2.Experiments with frontier proprietary and opensource multimodal models reveal persistent weaknesses in fine-grained visual-temporal perception, outcome verification, and error recovery, highlighting SWITCH as a testbed for closed-loop embodied intelligence.

cs.CV↗
arXiv2026

Executable verification through formalized expert reasoning in astronomical spectroscopy

Artificial intelligence has reshaped scientific prediction, but scientific verification remains a human bottleneck. Automated systems can map observations to labels, parameters or hypotheses, yet scientific conclusions require evidence, must satisfy physical consistency, and need explicit testing of alternatives before a decision is made. Here we introduce FORMA (Formalized Observational Reasoning with Auditable Decisions), an executable verification protocol that reconstructs expert reasoning into a workflow: it extracts evidence, generates hypotheses under physical constraints, tests alternatives, and performs auditable consistency checks. Unlike prediction or post-hoc interpretability, executable verification records and tests the evidential path leading to a decision. Astronomical spectroscopy provides a natural testbed, because ambiguous survey spectra are still adjudicated by expert visual inspection. Applied to the Dark Energy Spectroscopic Instrument (DESI) visual inspection catalogue, FORMA combines template-fitting candidate redshifts, spectral evidence extraction and physical audit into an auditable credibility score. A medium-or-higher credibility threshold identifies $331$ definite predictions with $95.5\%$ binary agreement with expert-adjudicated classes, while increasing credibility is associated with improved redshift consistency and higher classification reliability. These results show that automated inference can be coupled to explicit verification, allowing candidate outputs to be evaluated before they enter scientific use.

astro-ph.CO↗
arXiv2026

The Verification Horizon: No Silver Bullet for Coding Agent Rewards

A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, this intuition is being inverted: as foundation models develop stronger reasoning capabilities and engineering harnesses grow more sophisticated, generating complex candidate solutions is no longer difficult -- reliably verifying them has become the harder problem. Every verifier we can build is only a proxy for human intent, never the intent itself. This makes verification subject to a twofold difficulty: first, intent is underspecified by nature, making it inherently hard to faithfully check whether it has been fulfilled; second, during model training, optimization widens the gap between proxy and intent -- manifesting as reward hacking or signal saturation. To address this, we characterize the quality of verification signals along three dimensions -- scalability, faithfulness, and robustness -- and argue that achieving all three simultaneously is the central challenge. We further study four reward constructions: a test verifier for general coding tasks, a rubric verifier for frontend tasks, the user as verifier for real-world agent tasks, and an automated agent verifier for long-horizon tasks. Across different task types and policy capability levels, we conduct in-depth analysis and experiments on the core challenges of reward design and how to more effectively leverage reward signals. Experiments show that targeted verification design can effectively suppress reward hacking, improve task completion quality, and achieve significant gains across multiple internal and public benchmarks. These experiences collectively point to a core observation: no fixed reward function can remain effective as policy capability continues to grow; and verification must co-evolve with the generator.

cs.AI↗
arXiv2026

Semi-nonparametric estimation of spatial dynamic panel data models with nonparametric spatial weights

We develop a semi-nonparametric framework for spatial dynamic panel data (SDPD) models with two-way fixed effects when the spatial interaction structure is unknown beyond a distance measure. This is accomplished by modelling spatial weights in the outcome, lagged-outcome, and disturbance channels as unknown functions of underlying economic distances. These enter the SDPD system through matrix-function operators, providing a unified approach that accommodates both spatial autoregressive and matrix exponential spatial specifications. Allowing for unknown heteroskedasticity, we propose sieve GMM estimators based on a stacked set of linear and quadratic moment conditions, and derive a feasible optimal GMM estimator and a more efficient feasible best GMM estimator. As $(n, T) \rightarrow \infty$, the parametric component is $\sqrt{n(T - 1)}$-consistent and asymptotically normal, echoing classical semi-nonparametric results. Monte Carlo experiments indicate excellent finite-sample performance. We apply the method to 'witch' killings as studied by Miguel (2005), and find that economic-geography proximity rather than cultural-geography proximity between communities significantly amplifies spatial dependence in these economic murders.

econ.EM↗
arXiv2026

CulMind: Benchmarking Multimodal Understanding and Reasoning in Chinese Cultural Heritage

Evaluating Multimodal Large Language Models (MLLMs) in Chinese Cultural Heritage (CCH) requires fine-grained reasoning over visual, textual, stylistic, and historical clues. However, existing CCH benchmarks mainly emphasize final-answer accuracy, while the accuracy and completeness of reasoning processes remain underexplored. To address this gap, we introduce CulMind and CulMind-R: a high-quality benchmark for multimodal CCH covering 50 tasks from collections of more than 100 museums, and a 24-task reasoning subset that adaptively defines task-specific dimensions for reasoning process evaluation. To evaluate reasoning quality, we propose ReaScore, a task-adaptive metric that evaluates reasoning by automatically weighting task-relevant dimensions. Experiments on 14 leading MLLMs reveal a substantial gap between answers and reasoning, especially on challenging tasks. Further analysis shows that task-adaptive dimension selection and weighting better align evaluation results with expert judgments. Overall, our benchmark and metric support a more expert-aligned assessment of CCH understanding and offer a transferable reference for broader evaluations of cultural heritage. We publicly release the data, code, and evaluation scripts at https://github.com/ZevTsao/CulMind to facilitate reproducible research.

cs.CL↗
arXiv2026

VLM4VLA: Revisiting Vision-Language-Models in Vision-Language-Action Models

Vision-Language-Action (VLA) models, which integrate pretrained large Vision-Language Models (VLM) into their policy backbone, are gaining significant attention for their promising generalization capabilities. This paper revisits a fundamental yet seldom systematically studied question: how VLM choice and competence translate to downstream VLA policies performance? We introduce VLM4VLA, a minimal adaptation pipeline that converts general-purpose VLMs into VLA policies using only a small set of new learnable parameters for fair and efficient comparison. Despite its simplicity, VLM4VLA proves surprisingly competitive with more sophisticated network designs. Through extensive empirical studies on various downstream tasks across three benchmarks, we find that while VLM initialization offers a consistent benefit over training from scratch, a VLM's general capabilities are poor predictors of its downstream task performance. This challenges common assumptions, indicating that standard VLM competence is necessary but insufficient for effective embodied control. We further investigate the impact of specific embodied capabilities by fine-tuning VLMs on seven auxiliary embodied tasks (e.g., embodied QA, visual pointing, depth estimation). Contrary to intuition, improving a VLM's performance on specific embodied skills does not guarantee better downstream control performance. Finally, modality-level ablations identify the visual module in VLM, rather than the language component, as the primary performance bottleneck. We demonstrate that injecting control-relevant supervision into the vision encoder of the VLM yields consistent gains, even when the encoder remains frozen during downstream fine-tuning. This isolates a persistent domain gap between current VLM pretraining objectives and the requirements of embodied action-planning.

cs.CV↗
arXiv2026

Mechanistically Interpreting the Role of Sample Difficulty in RLVR for LLMs

Reinforcement Learning with Verifiable Reward (RLVR) is empirically shown to notably enhance the reasoning performance of large language models (LLMs), particularly in mathematics and programming. However, the mechanistic role of Sample Difficulty in RLVR remains poorly understood. In this paper, we investigate RLVR through the lens of difficulty-wise and one-sample analysis. We find that sample difficulty has a non-monotonic effect on RLVR: easy and medium-difficulty problems yield the strongest and most stable reasoning improvements, whereas overly hard problems often provide weak learning signals, induce degenerate behaviors such as answer repetition or skipping necessary computation, and can ultimately degrade the model's pre-existing capabilities. Beyond the obverse of response, we further analyze the model's internal feature dynamics using Temporal Sparse Autoencoders (T-SAE). Easy problems mainly reinforce direct-answer and basic-computation features while suppressing deliberative-reasoning features; hard problems activate reasoning-related features but become useful only when successful trajectories are sampled; medium-difficulty problems provide a more balanced signal, strengthening both computation and multi-step reasoning features. Motivated by these findings, we propose difficulty-adaptive strategies for hard-sample utilization, using backward-reasoning reformulation and T-SAE-guided training signals to improve reward density and credit assignment during RLVR. Overall, our results identify sample difficulty as a key factor governing both the optimization dynamics and representation evolution of RLVR.

cs.AI↗
arXiv2026

GAF: Gaussian Action Field as a 4D Representation for Dynamic World Modeling in Robotic Manipulation

Accurate scene perception is critical for vision-based robotic manipulation. Existing approaches typically follow either a Vision-to-Action (V-A) paradigm, predicting actions directly from visual inputs, or a Vision-to-3D-to-Action (V-3D-A) paradigm, leveraging intermediate 3D representations. However, these methods often struggle with action inaccuracies due to the complexity and dynamic nature of manipulation scenes. In this paper, we adopt a V-4D-A framework that enables direct action reasoning from motion-aware 4D representations via a Gaussian Action Field (GAF). GAF extends 3D Gaussian Splatting (3DGS) by incorporating learnable motion attributes, allowing 4D modeling of dynamic scenes and manipulation actions. To learn time-varying scene geometry and action-aware robot motion, GAF provides three interrelated outputs: reconstruction of the current scene, prediction of future frames, and estimation of init action via Gaussian motion. Furthermore, we employ an action-vision-aligned denoising framework, conditioned on a unified representation that combines the init action and the Gaussian perception, both generated by the GAF, to further obtain more precise actions. Extensive experiments demonstrate significant improvements, with GAF achieving +11.5385 dB PSNR, +0.3864 SSIM and -0.5574 LPIPS improvements in reconstruction quality, while boosting the average +7.3% success rate in robotic manipulation tasks over state-of-the-art methods.

cs.RO↗
arXiv2026

LENS: Multi-level Evaluation of Multimodal Reasoning with Large Language Models

Multimodal Large Language Models (MLLMs) have achieved significant advances in integrating visual and linguistic information, yet their ability to reason about complex and real-world scenarios remains limited. The existing benchmarks are usually constructed in the task-oriented manner without guarantee that different task samples come from the same data distribution, thus they often fall short in evaluating the synergistic effects of lower-level perceptual capabilities on higher-order reasoning. To lift this limitation, we contribute Lens, a multi-level benchmark with 3.4K contemporary images and 60K+ human-authored questions covering eight tasks and 12 daily scenarios, forming three progressive task tiers, i.e., perception, understanding, and reasoning. One feature is that each image is equipped with rich annotations for all tasks. Thus, this dataset intrinsically supports to evaluate MLLMs to handle image-invariable prompts, from basic perception to compositional reasoning. In addition, our images are manully collected from the social media, in which 53% were published later than Jan. 2025. We evaluate 15+ frontier MLLMs such as Qwen2.5-VL-72B, InternVL3-78B, GPT-4o and two reasoning models QVQ-72B-preview and Kimi-VL. These models are released later than Dec. 2024, and none of them achieve an accuracy greater than 60% in the reasoning tasks. Project page: https://github.com/Lens4MLLMs/lens. ICCV 2025 workshop page: https://lens4mllms.github.io/mars2-workshop-iccv2025/

cs.CV↗
arXiv2026

TokAlign++: Advancing Vocabulary Adaptation via Better Token Alignment

Tokenization is a foundational step in the text process of Large Language Models (LLMs). Texts must be first tokenized into token IDs, which are then input to LLMs. Inefficient tokenization results in long token-ID sequences and will slow down the training and inference of LLMs. The fine-grained knowledge transfer between LLMs, like token-level distillation, is also impeded by the mismatch in vocabulary. To bridge this gap, we introduce a method named TokAlign++ to improve vocabulary adaptation performance by learning better token alignment lexicon. The source and target vocabularies are taken as two different languages, and the bilingual token alignment lexicon is learned from monolingual token representations. Model parameters are rearranged following this bilingual lexicon for new vocabulary, and progressively fine-tuned for adaptation. Experimental results on 15 languages show that our method boosts the multilingual text compression rates and preserves most of the multilingual ability of vanilla models. It costs as few as 1k steps to restore the performance of the vanilla model. After unifying vocabularies between vanilla models, token-level distillation remarkably improves the base model with only 235M tokens.

cs.CL↗
arXiv2026

AdaMem: Adaptive User-Centric Memory for Long-Horizon Dialogue Agents

Large language model (LLM) agents increasingly rely on external memory to support long-horizon interaction, personalized assistance, and multi-step reasoning. However, existing memory systems still face three core challenges: they often rely too heavily on semantic similarity, which can miss evidence crucial for user-centric understanding; they frequently store related experiences as isolated fragments, weakening temporal and causal coherence; and they typically use static memory granularities that do not adapt well to the requirements of different questions. We propose AdaMem, an adaptive user-centric memory framework for long-horizon dialogue agents. AdaMem organizes dialogue history into working, episodic, persona, and graph memories, enabling the system to preserve recent context, structured long-term experiences, stable user traits, and relation-aware connections within a unified framework. At inference time, AdaMem first resolves the target participant, then builds a question-conditioned retrieval route that combines semantic retrieval with relation-aware graph expansion only when needed, and finally produces the answer through a role-specialized pipeline for evidence synthesis and response generation. We evaluate AdaMem on the LoCoMo and PERSONAMEM benchmarks for long-horizon reasoning and user modeling. Experimental results show that AdaMem achieves state-of-the-art performance on both benchmarks. The code will be released upon acceptance.

cs.CL↗
arXiv2026

World2VLM: Distilling World Model Imagination into VLMs for Dynamic Spatial Reasoning

Vision-language models (VLMs) have shown strong performance on static visual understanding, yet they still struggle with dynamic spatial reasoning that requires imagining how scenes evolve under egocentric motion. Recent efforts address this limitation either by scaling spatial supervision with synthetic data or by coupling VLMs with world models at inference time. However, the former often lacks explicit modeling of motion-conditioned state transitions, while the latter incurs substantial computational overhead. In this work, we propose World2VLM, a training framework that distills spatial imagination from a generative world model into a vision-language model. Given an initial observation and a parameterized camera trajectory, we use a view-consistent world model to synthesize geometrically aligned future views and derive structured supervision for both forward (action-to-outcome) and inverse (outcome-to-action) spatial reasoning. We post-train the VLM with a two-stage recipe on a compact dataset generated by this pipeline and evaluate it on multiple spatial reasoning benchmarks. World2VLM delivers consistent improvements over the base model across diverse benchmarks, including SAT-Real, SAT-Synthesized, VSI-Bench, and MindCube. It also outperforms the test-time world-model-coupled methods while eliminating the need for expensive inference-time generation. Our results suggest that world models can serve not only as inference-time tools, but also as effective training-time teachers, enabling VLMs to internalize spatial imagination in a scalable and efficient manner.

cs.CV↗