Pith. sign in

REVIEW 3 major objections 7 minor 19 references

Separating causal from spurious features and weighting class prototypes by reliability lets federated models keep learning when feature distributions change across clients and over time, reaching the highest average accuracy on three domain

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 17:35 UTC pith:ULDYJVWB

load-bearing objection Solid FL engineering package with real ablations and modest SOTA gains, but the “dynamic causal” framing is not what the experiments test. the 3 major comments →

arxiv 2607.09695 v1 pith:ULDYJVWB submitted 2026-06-20 cs.LG

FedCausal-Dyn: A Causal-Dynamic Paradigm for Federated Learning under Dynamic Feature Drift

Kaijie Chen , Alex Johnson , Maria Garcia , Wei Zhang , Daniel Kim This is my paper
classification cs.LG
keywords federated learningdynamic feature driftcausal feature separationprototype aggregationdomain generalizationadversarial trainingnon-stationary data
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Federated learning usually keeps raw data on each client, but when the same class looks different across clients and those differences keep changing over time, the shared model degrades. This paper claims that treating the problem as dynamic feature drift—not static non-IID data—and solving it with a causal-dynamic pipeline restores accuracy and stability. The method first peels domain-invariant causal features away from client-specific spurious ones with two projection heads and adversarial training, then builds global class prototypes only from the causal stream and weights each client’s prototype by how reliable that client looks. A single collaborative regularizer then pulls local causal features toward those reliable global prototypes while still enforcing domain invariance. On Office-10, Digits, and PACS the resulting system reports the best average accuracies and the lowest run-to-run variance among a large set of federated and domain-generalization baselines, with ablations showing each piece contributes.

Core claim

The paper establishes that dynamic feature drift in federated learning can be handled by explicitly disentangling domain-invariant causal features from domain-specific spurious features, then aggregating only the causal class prototypes with reliability-aware weights and regularizing local models against those prototypes. When this causal-dynamic loop is closed, the global model achieves state-of-the-art average top-1 accuracy (68.95 % Office-10, 89.34 % Digits, 61.26 % PACS) and the most stable results across three standard federated domain-generalization benchmarks.

What carries the argument

Causal-domain feature separation: two lightweight projection heads on a shared encoder, trained with causal prediction, domain-invariance (via gradient reversal), and domain-capturing losses so that only the causal stream is used for reliability-weighted prototype aggregation and contrastive alignment.

Load-bearing premise

The method assumes that a pair of projection heads plus domain-adversarial training can cleanly separate the features that truly predict the label from the features that only mark which client the data came from; if that separation fails, the prototypes and regularizer target the wrong representation.

What would settle it

Re-run the three benchmarks with the causal and spurious heads forcibly collapsed into a single adversarial representation (or with deliberately entangled synthetic factors); if the reported accuracy and stability gaps over the strongest baselines disappear, the central claim that the separation is what drives the gains is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Federated systems facing non-stationary streams (e.g., changing market or sensor conditions) can keep a usable global model without sharing raw features or synthetic data.
  • Reliability-weighted prototype aggregation supplies a practical filter against low-quality or temporarily unreliable client updates.
  • The same causal-stream prototypes can serve both inter-client alignment and domain-invariance, reducing the need for separate ad-hoc regularizers.
  • Ablation patterns imply that removing the collaborative regularizer hurts more than removing any other single component, so future designs should treat prototype contrastive alignment as primary.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The reliability weight itself could be treated as a first-class federated signal and reused for client selection or early stopping, even outside the causal-separation setting.
  • If the separation works under true concept drift rather than only style/domain shift, the same pipeline may transfer to streaming recommendation or fraud models where both features and labels slowly change.
  • The privacy-enhanced mixup step is largely orthogonal; it could be swapped for stronger cryptographic or differential-privacy mechanisms without rewriting the causal core.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. The manuscript proposes FedCausal-Dyn, a federated learning framework for dynamic feature drift. It models client data as time-evolving domains D_n^t (§3.1) and introduces three components: (i) causal–domain feature separation via dual projection heads C(·), S(·) and adversarial losses L_Causal, L_Inv, L_Spur (Eqs. 2–4); (ii) reliability-weighted dynamic prototype aggregation (Eqs. 5–6); and (iii) causal-feature guided collaborative regularization L_Reg combining prototype contrastive alignment with domain invariance (Eq. 7), plus a privacy-enhanced causal mixup (Eqs. 8–9). Experiments on Office-10, Digits, and PACS report state-of-the-art average accuracies (68.95%, 89.34%, 61.26%) and lowest average standard deviation versus a broad baseline suite (Tables 1–4), with ablations (Tables 5–8) attributing the largest gains to L_Reg and supporting each module.

Significance. If the method genuinely improved federated learning under temporal non-stationarity, the contribution would be practically important for settings such as financial technology that the abstract highlights. The empirical package is carefully executed for static multi-domain federated domain generalization: three seeds, many baselines, component ablations, reliability-weight variants, and λ sensitivity. Those results would still be a useful incremental advance over prototype and adversarial FL methods (FedProto, ADCOL, MOON, FedBN). The significance of the stated “causal-dynamic” claim, however, is limited by the gap between the temporal formulation and the static evaluation protocol; closing that gap (or reframing) is what would make the paper’s central story load-bearing rather than aspirational.

major comments (3)
  1. §3.1 formulates the problem as dynamic feature drift: domains D_n^t with non-stationary P_n^t(x|y) over communication rounds t, and Eq. (1) explicitly takes an expectation over t. The abstract and introduction contrast this with prior work that “assume[s] static drift.” Yet §4 evaluates only the standard static multi-domain leave-one-domain-out protocol on Office-10, Digits, and PACS; no client’s P(x|y) is altered across rounds. Consequently the “dynamic” update of g_k^t (Eq. 6) and the reliability weights ω_n,k^t never face genuine temporal concept drift. The SOTA averages in Tables 1–4 therefore do not substantiate the paper’s central claim as stated. Either inject controlled temporal drift (e.g., rotating domain statistics or gradual style shifts across rounds) and re-evaluate, or reframe the contribution as federated domain generalization / static feature drift and revise the abstrac
  2. §3.2.1 and Eqs. (2)–(4) treat the dual-head adversarial construction as separating domain-invariant causal features z_c from spurious z_s. No identification argument, intervention, or causal metric is provided; success is inferred from held-out accuracy and a qualitative t-SNE case study (§6.2). The paper itself notes in §7 that separation “may be less effective” when factors are highly entangled. If z_c is only partially purified, reliability-weighted prototypes and L_Reg (Eq. 7) align to the wrong representation, so the mechanistic explanation of the gains in Tables 1–5 does not follow. At minimum, add quantitative checks (domain predictability of z_c vs z_s, label predictability of each stream, or controlled spurious-correlation synthetic data) so that the “causal” label is empirically justified rather than terminological.
  3. Table 4 and §4.2.4 claim overall rank 1.00 and “most stable results,” but the experimental suite never includes a true non-stationary baseline or a dynamic-drift ablation of the reliability weights. Without that, the attribution of stability to “reliable and dynamic prototype aggregation” under dynamic drift (§4.2.5, §6.1) remains untested. A minimal load-bearing addition is a controlled non-stationarity experiment (or an explicit statement that all reported gains are under static multi-domain heterogeneity only).
minor comments (7)
  1. Related Work and Introduction cite a large number of concurrent/arXiv works (including many from overlapping author groups) that are only loosely connected to feature drift; tighten to the methods that actually compete on the same problem (FedProto, ADCOL, MOON, FedBN, FRAug, FedCiR).
  2. §3.2.2: reliability weight ω is defined as either validation accuracy or inverse intra-class variance, but the main tables do not state which was used for the reported SOTA numbers; Table 6 shows Acc is best—make the default explicit in §4.1.
  3. Eq. (9) is a constrained min–max over a binary mask m; implementation details (how m is optimized, whether it is continuous relaxation, cost per round) are missing and should be added for reproducibility.
  4. Figure 1 is described but the manuscript text does not specify encoder backbone, local epochs, client sampling rate, or communication rounds used for Tables 1–3; add a hyperparameter table.
  5. Notation: D_n^t is used for both the domain and the local dataset; distinguish distribution vs. sample set. Also L_Inv appears both as a standalone adversarial loss (Eq. 3) and inside L_Reg (Eq. 7)—clarify whether the same discriminator is shared.
  6. §6.3 lists “potential extended analyses” (partial participation, label noise) without results; either run them or move to future work to avoid implying unfinished experiments.
  7. Typos/style: “FedCausal-Dyn” sometimes loses bold/spacing; “PerfedAvg” vs “Per-FedAvg”; duplicate FedProto citations (Tan et al. 2022a/b/c).

Circularity Check

1 steps flagged

Mild motivational self-citation of graph-diagnosability papers; method and SOTA claims remain independently evaluated on held-out accuracy.

specific steps
  1. self citation load bearing [§3.1 (Problem Formulation) and §3.2.2 (Reliable Aggregation)]
    "These insights conceptually support our formulation of dynamic feature drift and motivate the need for principled mechanisms to maintain invariant global representations under evolving local conditions Wang et al. [2017], Lin et al. [2017], Wang et al. [2018], Li et al. [2022], Xiang et al. [2025], Wang et al. [2013], Niu et al. [2024b]. ... These theoretical insights motivate our design of reliability-weighted prototype aggregation ... Wang et al. [2020], Wang and Wang [2021], Wang et al. [2025a], Xiang et al. [2025], Pan et al. [2024], Xu et al. [2025]."

    The paper's central framing of 'dynamic feature drift' and the reliability-weighting scheme are justified solely by citations whose author lists heavily overlap with the present paper (Wang, Lin, Xiang, Niu, etc.). The cited results concern restricted connectivity and diagnosability of interconnection networks, not federated learning; they supply no independent theorem that forces the FL design. The citations therefore function as self-referential motivation rather than external evidence, though they remain non-load-bearing for the accuracy numbers.

full rationale

The paper's core derivation is an engineering construction (projection heads + GRL adversarial losses L_Causal/L_Inv/L_Spur, reliability-weighted prototype mean Eq. 6, contrastive L_Reg) whose success is measured by ordinary top-1 accuracy on three static domain-generalization benchmarks. None of the reported numbers is forced by construction from a fitted parameter or from an identity that equates input to output. The only circularity-adjacent pattern is repeated citation of the authors' own (or closely overlapping) graph-connectivity/diagnosability results as 'conceptual support' or 'motivation' for the dynamic-drift formulation and reliability weighting; those citations are not load-bearing for the empirical claims and do not import a uniqueness theorem that forbids alternatives. Experiments, ablations, and baselines are self-contained against external data, so the circularity score remains low.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 3 invented entities

The central SOTA claim rests on modeling choices and free knobs rather than parameter-free theory. Load-bearing modeling axioms are the causal/spurious split, shared label prior, and the validity of local reliability scores as aggregation weights. Free parameters (λ, γ, η, τ, reliability definition, mask constraint ε) are tuned or chosen and affect reported accuracy. Invented entities are the dual extractors, reliability-weighted global prototypes, and the uncertainty-maximizing mixup mask—engineering constructs without independent external evidence beyond the paper’s own ablations.

free parameters (5)
  • λ (domain-invariance weight in L_Reg)
    Balances L_Align vs L_Inv; Table 7 shows sensitivity and selects λ=0.5 as best/near-best. Directly affects the main accuracy numbers.
  • γ, η (weights on L_Reg and L_Spur in L_Local)
    Eq. 10 introduces γ and η as hyperparameters with no derivation; they control how much collaborative regularization and spurious capture enter local training.
  • τ (contrastive temperature)
    Appears in the InfoNCE-style L_Align (Eq. 7); standard free temperature that shapes the prototype alignment geometry.
  • Reliability weight ω_n,k definition (Acc vs Con vs Uniform)
    Table 6 compares strategies; Accuracy-based is chosen for the full model. The aggregation formula (Eq. 6) depends on this design choice, not on a fixed physical constant.
  • Mixup mask constraint ε and domain-entropy objective
    Eqs. 8–9 introduce a min-max privacy mixup with similarity floor ε; values are not derived from first principles and affect the global classifier training signal.
axioms (5)
  • ad hoc to paper Encoder features z = G(x) decompose into domain-invariant causal features z_c predictive of y and domain-specific spurious features z_s that can be isolated by two projection heads plus adversarial domain losses.
    Stated as a posit in §3.2.1; operationalized by C(·), S(·), L_Causal, L_Inv, L_Spur. Not proven; limitations (§7) admit failure under entanglement.
  • domain assumption Feature drift is variation in P_n^t(x|y) across clients and rounds while the label prior P(y) is shared.
    Problem formulation §3.1; standard FL/DG modeling choice that excludes label-shift-dominated regimes.
  • domain assumption Local validation accuracy or inverse intra-class variance is a valid reliability weight for prototype aggregation under dynamic/heterogeneous clients.
    §3.2.2 and Table 6; required for Eq. 6 to improve on uniform averaging. Assumes held-out data or stable variance estimates exist.
  • domain assumption Gradient-reversal adversarial training against a client-ID discriminator removes domain information from z_c without destroying label-predictive content.
    Eq. 3 cites GRL/DANN-style training; standard domain-adaptation assumption imported into the FL setting.
  • standard math Standard cross-entropy, cosine-similarity contrastive (InfoNCE-style), and federated multi-round optimization behave as usual under partial client participation A_t.
    Background optimization and loss machinery used throughout §3 without new theorems.
invented entities (3)
  • Causal Feature Extractor C(·) and Spurious Feature Extractor S(·) no independent evidence
    purpose: Implement the claimed causal–domain separation that feeds prototypes and regularization.
    New architectural split relative to single-representation FL baselines; evidence is only internal ablations (Table 5, 7), not external causal tests.
  • Reliability-weighted dynamic global prototypes g_k^t with weights ω_n,k^t no independent evidence
    purpose: Replace sample-count averaging so “reliable” clients dominate the global class representation.
    Defined in Eqs. 5–6; quality depends on the chosen reliability estimator. No external benchmark of the weights themselves.
  • Privacy-enhanced causal-feature mixup with uncertainty-maximizing binary mask m no independent evidence
    purpose: Upload mixed causal features that preserve label utility while maximizing domain-discriminator entropy.
    Eqs. 8–9; mini-max construction local to this paper. No formal privacy bound (ε-DP etc.) is proven.

pith-pipeline@v1.1.0-grok45 · 22106 in / 4378 out tokens · 47098 ms · 2026-07-14T17:35:14.162339+00:00 · methodology

0 comments
read the original abstract

This paper addresses the challenging problem of dynamic feature drift in federated learning, where data distributions evolve across clients and over time -- a common scenario in real-world applications like financial technology. Existing approaches often assume static drift, limiting their effectiveness in non-stationary environments. To overcome this, we propose \textbf{FedCausal-Dyn}, a novel federated learning framework built on a causal-dynamic paradigm. Its key innovation is \textit{causal-domain feature separation}, which disentangles domain-invariant causal features from spurious, domain-specific variations via specialized projection heads and adversarial training. This enables \textit{reliable and dynamic prototype aggregation}, weighting local class prototypes by estimated reliability before global aggregation. We further introduce \textit{causal-feature guided collaborative regularization}, unifying prototype contrastive alignment and domain invariance into a cohesive objective. Extensive experiments on three federated domain generalization benchmarks demonstrate that FedCausal-Dyn consistently achieves state-of-the-art performance, with the highest average accuracy and the most stable results. Ablation studies confirm each component's critical contribution. Our work provides a robust and principled solution for federated learning under dynamic feature drift.

Figures

Figures reproduced from arXiv: 2607.09695 by Alex Johnson, Daniel Kim, Kaijie Chen, Maria Garcia, Wei Zhang.

Figure 1
Figure 1. Figure 1: Overview of the FedCausal-Dyn framework. Each client performs causal–domain feature [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Ablation study visualization showing the contribution of each component across three [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Heatmap showing the performance contribution of each component (C: Causal-Domain [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Sensitivity analysis of the regularization balance hyperparameter [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Training curves showing loss and accuracy progression across communication rounds for [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

19 extracted references · 9 linked inside Pith

  1. [1]

    Multi-agent collaborative framework for intelligent it operations: An aoi system with context-aware compression and dynamic task scheduling.arXiv preprint arXiv:2512.13956,

    Zishan Bai, Enze Ge, and Junfeng Hao. Multi-agent collaborative framework for intelligent it operations: An aoi system with context-aware compression and dynamic task scheduling.arXiv preprint arXiv:2512.13956,

    Pith/arXiv arXiv

  2. [2]

    Exploring efficiency frontiers of thinking budget in medical reasoning: Scaling laws between computational resources and reasoning quality.arXiv:2508.12140,

    Ziqian Bi, Lu Chen, Junhao Song, Hongying Luo, Enze Ge, Junmin Huang, Tianyang Wang, Keyu Chen, Chia Xin Liang, Zihan Wei, et al. Exploring efficiency frontiers of thinking budget in medical reasoning: Scaling laws between computational resources and reasoning quality.arXiv:2508.12140,

    arXiv

  3. [3]

    Exploiting shared represen- tations for personalized federated learning

    13 Liam Collins, Hamed Hassani, Aryan Mokhtari, and Sanjay Shakkottai. Exploiting shared represen- tations for personalized federated learning. InInternational Conference on Machine Learning, pages 2089–2099. PMLR,

    2089

  4. [4]

    Graph inference towards icd coding.arXiv preprint arXiv:2601.07496,

    Xiaoxiao Deng. Graph inference towards icd coding.arXiv preprint arXiv:2601.07496,

    arXiv

  5. [5]

    Plan then action: High-level planning guidance reinforcement learning for llm reasoning.arXiv preprint arXiv:2510.01833,

    Zhihao Dou, Qinjian Zhao, Zhongwei Wan, Dinggen Zhang, Weida Wang, Towsif Raiyan, Benteng Chen, Qingtao Pan, Yang Ouyang, Zhiqiang Gao, et al. Plan then action: High-level planning guidance reinforcement learning for llm reasoning.arXiv preprint arXiv:2510.01833,

    Pith/arXiv arXiv

  6. [6]

    Dsadf: Thinking fast and slow for decision making.International Journal of Computer Vision, 134(6):270, 2026a

    Zhihao Dou, Dongfei Cui, Jun Yan, Weida Wang, Benteng Chen, Haoming Wang, Zeke Xie, and Shufei Zhang. Dsadf: Thinking fast and slow for decision making.International Journal of Computer Vision, 134(6):270, 2026a. Zhihao Dou, Qinjian Zhao, Zhongwei Wan, Xiaoyu Xia, and Sumon Biswas. Core-code: Collabora- tive reinforcement learning for code generation.arXi...

    Pith/arXiv arXiv

  7. [7]

    Multi-agent medical decision consensus matrix system: An intelligent collaborative framework for oncology mdt consultations.arXiv preprint arXiv:2512.14321,

    Xudong Han, Xianglun Gao, Xiaoyi Qu, and Zhenyu Yu. Multi-agent medical decision consensus matrix system: An intelligent collaborative framework for oncology mdt consultations.arXiv preprint arXiv:2512.14321,

    arXiv

  8. [8]

    Ge-adapter: A general and efficient adapter for enhanced video editing with pretrained text-to-image diffusion models.Expert Systems with Applications, page 129649, 2025a

    Yangfan He, Sida Li, Kun Li, Jianhui Wang, Binxu Li, Tianyu Shi, Yi Xin, Keqin Li, Jun Yin, Miao Zhang, et al. Ge-adapter: A general and efficient adapter for enhanced video editing with pretrained text-to-image diffusion models.Expert Systems with Applications, page 129649, 2025a. Yangfan He, Sida Li, Jianhui Wang, Kun Li, Xinyuan Song, Xinhang Yuan, Keq...

    Pith/arXiv arXiv

  9. [9]

    Abductive inference in retrieval-augmented language models: Generating and validating missing premises, 2025a

    Shiyin Lin. Abductive inference in retrieval-augmented language models: Generating and validating missing premises, 2025a. URLhttps://arxiv.org/abs/2511.04020. Shiyin Lin. Hybrid fuzzing with llm-guided input mutation and semantic feedback, 2025b. URL https://arxiv.org/abs/2511.03995. Shiyin Lin. Llm-driven adaptive source-sink identification and false po...

    arXiv

  10. [10]

    From text to multimodality: Exploring the evolution and impact of large language models in medical practice, 2024a

    Qian Niu, Keyu Chen, Ming Li, Pohsun Feng, Ziqian Bi, Lawrence KQ Yan, Yichao Zhang, Cait- lyn Heqi Yin, Cheng Fei, Junyu Liu, Benji Peng, Tianyang Wang, Yunze Wang, Silin Chen, and Ming Liu. From text to multimodality: Exploring the evolution and impact of large language models in medical practice, 2024a. URLhttps://arxiv.org/abs/2410.01812. Qian Niu, Ju...

    arXiv

  11. [11]

    Deep learning and machine learning: Contrastive learning, from scratch to application

    Xinyuan Song, Keyu Chen, Ziqian Bi, Qian Niu, Junyu Liu, Benji Peng, Sen Zhang, Ming Liu, Ming Li, Xuanhe Pan, et al. Deep learning and machine learning: Contrastive learning, from scratch to application. 2025a. 15 Xinyuan Song, Yangfan He, Sida Li, Jianhui Wang, Hongyang He, Xinhang Yuan, Ruoyu Wang, Jiaqi Chen, Keqin Li, Kuan Lu, et al. Efficient tempor...

    Pith/arXiv arXiv

  12. [12]

    Huanhuan Wang, Xiao Zhang, Youbing Xia, and Xiang Wu

    URLhttps://arxiv.org/abs/2511.01243. Huanhuan Wang, Xiao Zhang, Youbing Xia, and Xiang Wu. An intelligent blockchain-based access control framework with federated learning for genome-wide association studies.Computer Standards & Interfaces, 84:103694,

    arXiv

  13. [13]

    Deep learning model security: Threats and defenses

    Tianyang Wang, Ziqian Bi, Yichao Zhang, Ming Liu, Weiche Hsieh, Pohsun Feng, Lawrence KQ Yan, Yizhu Wen, Benji Peng, Junyu Liu, et al. Deep learning model security: Threats and defenses. InarXiv:2412.08969, 2024b. Tianyang Wang, Ming Liu, Benji Peng, Xinyuan Song, Charles Zhang, Xintian Sun, Qian Niu, Junyu Liu, Silin Chen, Keyu Chen, Ming Li, Pohsun Feng...

    Pith/arXiv arXiv

  14. [14]

    Fstgat: Financial spatio-temporal graph attention network for non-stationary financial systems and its application in stock price prediction.Symmetry, 17(8):1344, 2025a

    Ze-Lin Wei, Hong-Yu An, Yao Yao, Wei-Cong Su, Guo Li, Saifullah, Bi-Feng Sun, and Mu-Jiang- Shan Wang. Fstgat: Financial spatio-temporal graph attention network for non-stationary financial systems and its application in stock price prediction.Symmetry, 17(8):1344, 2025a. Zhang Wei, Peilu Hu, Shengning Lang, Hao Yan, Li Mei, Yichao Zhang, Chen Yang, Junfe...

    Pith/arXiv arXiv

  15. [15]

    Lumina-dimoo: An omni diffusion large language model for multi-modal generation and understanding.arXiv preprint arXiv:2510.06308, 2025a

    Yi Xin, Qi Qin, Siqi Luo, Kaiwen Zhu, Juncheng Yan, Yan Tai, Jiayi Lei, Yuewen Cao, Keqi Wang, Yibin Wang, et al. Lumina-dimoo: An omni diffusion large language model for multi-modal generation and understanding.arXiv preprint arXiv:2510.06308, 2025a. Yi Xin, Juncheng Yan, Qi Qin, Zhen Li, Dongyang Liu, Shicheng Li, Victor Shea-Jay Huang, Yupeng Zhou, Ren...

    arXiv

  16. [16]

    Chen Yang, Yangfan He, Aaron Xuxiang Tian, Dong Chen, Jianhui Wang, Tianyu Shi, Arsalan Heydarian, and Pei Liu

    URLhttps://arxiv.org/abs/2410.21348. Chen Yang, Yangfan He, Aaron Xuxiang Tian, Dong Chen, Jianhui Wang, Tianyu Shi, Arsalan Heydarian, and Pei Liu. Wcdt: World-centric diffusion transformer for traffic scene generation. In2025 IEEE International Conference on Robotics and Automation (ICRA), pages 6566–6572. IEEE, 2025a. Junjie Yang, Junhao Song, Xudong H...

    arXiv

  17. [17]

    Af- fective multimodal agents with proactive knowledge grounding for emotionally aligned marketing dialogue.arXiv preprint arXiv:2511.21728, 2025a

    Lin Yu, Xiaofei Han, Yifei Kang, Chiung-Yi Tseng, Danyang Zhang, Ziqian Bi, and Zhimo Han. Af- fective multimodal agents with proactive knowledge grounding for emotionally aligned marketing dialogue.arXiv preprint arXiv:2511.21728, 2025a. Zhenyu Yu. Ai for science: A comprehensive review on innovations, challenges, and future directions. International Jou...

    arXiv

  18. [18]

    Qinjian Zhao, Zhihao Dou, Dinggen Zhang, Xiangyu Li, Chaoda Song, Zhongwei Wan, Xinpeng Li, Yanyan Zhang, Kaijie Chen, Qingtao Pan, et al

    URL https://arxiv.org/abs/2503.13522. Qinjian Zhao, Zhihao Dou, Dinggen Zhang, Xiangyu Li, Chaoda Song, Zhongwei Wan, Xinpeng Li, Yanyan Zhang, Kaijie Chen, Qingtao Pan, et al. Stride: Strategic trajectory reasoning via discriminative estimation for verifiable reinforcement learning.arXiv preprint arXiv:2606.15866,

    Pith/arXiv arXiv

  19. [19]

    Reagent-v: A reward-driven multi-agent framework for video understanding.arXiv preprint arXiv:2506.01300,

    Yiyang Zhou, Yangfan He, Yaofeng Su, Siwei Han, Joel Jang, Gedas Bertasius, Mohit Bansal, and Huaxiu Yao. Reagent-v: A reward-driven multi-agent framework for video understanding.arXiv preprint arXiv:2506.01300,

    Pith/arXiv arXiv