Exploiting Overlapping Fields of View for Redundancy-Aware Uplink Transmission in Vehicular 6G

Hamidreza Mazandarani; Masoud Shokrnezhad; Onur G\"unl\"u; Tarik Taleb

REVIEW 2 major objections 5 minor 18 references

Exploiting Overlapping Fields of View for Redundancy-Aware Uplink Transmission in Vehicular 6G

T0 review · 2 major / 5 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read Vehicles with overlapping views can skip sending redundant image patches and still reconstruct high-fidelity streams at the base station.

desk verdict Solid systems framing of joint patch selection + MAC for FoV-overlapping vehicular uplink; gains are real under oracle reconstruction but the abstract claim overreaches what Fig. 4 actually measures. read the letter →

arxiv 2606.31715 v2 pith:G6NHKGUG submitted 2026-06-30 cs.NI cs.CV

Hamidreza Mazandarani , Masoud Shokrnezhad , Tarik Taleb , Onur Günlü This is my paper · ORCID

classification cs.NIcs.CV

keywords 6GsemanticcommunicationsvehicularnetworksmultipleaccessspatialredundancyV2Vfieldofviewuplink

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Uplink-heavy 6G services such as cooperative vehicular streaming must move large volumes of camera data over scarce wireless spectrum. Standard semantic communication methods already drop task-irrelevant content, but they still treat each vehicle as an independent source and therefore retransmit the same scene elements many times when fields of view overlap. This paper claims that a multiple-access scheme can treat spectrum allocation as joint perception and transmission control: vehicles first exchange a few carefully chosen patches over short-range V2V links, compute which patches are spatially redundant, and then uplink only the unique, semantically important patches. The base station reconstructs each vehicle’s full observation by combining those patches with complementary views from neighbors. In a dense CARLA urban simulation the method raises the fraction of users that meet high-fidelity reconstruction thresholds, especially when channel resources are moderate. The practical payoff is fewer redundant bits on the air while still supporting applications that need high visual quality.

What carries the argument

Two-phase procedure: (1) FoV-priority V2V patch sharing so each vehicle can compute spatial redundancy via similarity matching (SSIM or embedding cosine + Hungarian assignment); (2) importance-weighted scheduling of only the remaining unique patches on V2I channels, with the base station later reconstructing missing patches from neighbors’ complementary views.

What would settle it

Replace the perfect-view-transformation oracle with a practical multi-view synthesizer or diffusion inpainting module on the same CARLA traces and measure whether the fraction of users meeting the SSIM/LPIPS quality threshold still exceeds the importance-aware baseline under the same moderate channel budgets.

Watch

Extended reading notes

Core claim

Exploiting partial field-of-view overlap among nearby vehicles lets a two-phase semantic multiple-access scheme transmit only non-redundant image patches, and thereby increases the proportion of users that achieve high-fidelity reconstruction under limited uplink resources, relative to importance-aware baselines that ignore inter-user redundancy.

Load-bearing premise

The base station can reconstruct missing patches to high fidelity from neighboring vehicles’ complementary views; the reported gains rest on a perfect-view-transformation upper bound rather than a real reconstructor.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The useful core is a clean joint formulation of perception and multiple access for dense vehicular camera uplink, plus a practical two-phase pipeline: short V2V exchange to estimate spatial redundancy, then semantic V2I of only non-redundant patches. That is more concrete than most “semantic MAC” papers that still treat users independently. Constraints (single-channel hardware, collision cliques, at-most-once patch transmission) are stated carefully, and the CARLA setup (40 vehicles, 10×10 patches, V2V/V2I split) is realistic enough for the subfield.

What is new is the combination: FoV-priority ordering for observation sharing, Hungarian matching on SSIM/JEPA-style similarity, and the importance weight ω that folds object coverage, depth, and cross-user reconstruction utility. The unimodal trade-off on observation-sharing duration is a sensible systems result, and the moderate-resource regime is exactly where redundancy exploitation should matter. Citations cover correlated sources, sensor coverage, and recent semantic token/importance work without obvious gaps.

The soft spot is load-bearing and exactly as the stress-test says. Objective (P) and reconstruction (8) define success by high-fidelity Q after multi-view synthesis of missing patches. Evaluation never runs a reconstructor. “Ours UB” assumes perfect view transformation (oracle access to the target view); LB reuses the raw preceding frame. Practical modules (LoFTR, GenAI inpainting) are discussed then deferred. So Fig. 4 shows that non-redundant selection helps under idealized reconstruction, not that the end-to-end high-fidelity claim holds once warping artifacts, dynamics, or occlusions appear. Free knobs (β, γ, η, D_OS, role thresholds) and missing scheduler/code details further limit reproducibility. These are fixable with an actual reconstructor and fuller baselines, not fatal to the idea.

This is for people working on multi-user semantic uplink or cooperative vehicular perception who need a concrete MAC+perception formulation. It deserves a serious referee; the framing and simulation design are already good enough that the reconstruction gap should be fixed in revision rather than used for desk rejection. I would engage: cite the problem statement and two-phase idea, and watch for the follow-up that closes the oracle gap.

Referee Report

2 major / 5 minor

Summary. The paper proposes a semantic-aware multiple-access scheme for uplink-dominant vehicular 6G streaming that exploits overlapping fields of view among nearby vehicles. It formulates a joint perception–transmission problem (P) in which users decide which image patches to send, when, and over which channel under bandwidth, single-channel hardware, and collision constraints. A practical two-phase algorithm is given: V2V observation sharing (guided by a geometric FoV-priority heuristic) to estimate inter-user redundancy via patch matching, followed by BS-scheduled V2I transmission of only non-redundant, semantically important patches (importance ω in Eq. (9)). At the BS, missing patches are to be recovered by multi-view synthesis / GenAI reconstruction (Eq. (8)). CARLA simulations with 40 vehicles claim that the scheme raises the fraction of users meeting SSIM/LPIPS quality thresholds relative to an importance-aware semantic baseline, especially under moderate channel resources (Fig. 4).

Significance. If the end-to-end claim holds, the work is a useful step beyond single-user or independent multi-user semantic MAC: it treats spectrum access as joint perception and transmission control and shows how modest V2V cooperation can free uplink resources in dense vehicular settings. The problem statement is carefully constrained (Eqs. (2)–(6)), the two-phase decomposition is implementable, and the simulation includes realistic hardware and collision limits. The main contribution is therefore architectural and empirical rather than theoretical; its value for sustainable 6G uplink depends on whether the reported gains survive practical multi-view reconstruction rather than an oracle upper bound.

major comments (2)

The central claim (Abstract; objective (P); reconstruction (8); §IV, Fig. 4) is that the scheme improves the proportion of users achieving high-fidelity reconstruction. Evaluation, however, never runs a practical reconstructor. “Ours UB” assumes perfect view transformation of the preceding vehicle’s observation into the following vehicle’s viewpoint (explicitly “equivalent to directly accessing the following user’s observation”), while LB reuses the raw preceding view. Practical modules (LoFTR warping, GenAI inpainting, multimodal models) are only discussed in §III-B and deferred to future work (§V). Consequently the plotted SSIM/LPIPS satisfaction ratios measure the value of non-redundant patch selection under an idealized oracle, not under the reconstruction process the abstract and problem statement invoke. Residual holes, warping artifacts, or hallucinations on dynamic/occluded conte
Free parameters that directly shape the reported gains are not ablated or justified beyond hand-chosen values: β, γ in the importance score ω (Eq. (9)), similarity threshold η, observation-sharing duration D_OS (shown unimodal in Fig. 4 bottom but not systematically optimized), and the preceding/following distance and orientation thresholds (e.g., 50 m, 90°). Because the largest gains appear precisely in the moderate-resource regime where these knobs matter most, the paper should report sensitivity of the user-satisfaction ratios to these choices, or fix them by a clear rule, so that the improvement over SOTA is not an artifact of tuning.

minor comments (5)

Title on the arXiv page (“Exploiting Overlapping Fields of View…”) differs slightly from the manuscript title (“Semantic-Aware Multiple Access via Spatial Redundancy Exploitation…”); align them.
Fig. 1 caption and body text refer to “highlighted 4th bars” and “blue patch” without a clear visual legend; add labels or a color key.
Notation for time frames vs. slots (t vs. τ, T^t) is dense; a short table of symbols would help.
SOTA baseline is cited as “[3]” (OFDM-based digital semantic communication); briefly state which components of that work are retained so the comparison is reproducible.
Typographical inconsistencies: “V AEs”, “G ünlü”, and occasional missing spaces around citations.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: empirical CARLA comparison under stated bounds, not a result forced by definition or self-citation.

full rationale

The paper is a systems proposal with a joint perception/MAC formulation (P) and a practical two-phase heuristic (V2V sharing then non-redundant V2I uplink). The central claim is an empirical simulation outcome: under limited channels, selective non-redundant patch transmission raises the fraction of users meeting SSIM/LPIPS thresholds relative to an importance-aware SOTA baseline (Fig. 4). That outcome is measured from CARLA traffic, patch selection, and scheduling; it is not algebraically identical to the inputs. Equation (9) uses tunable β, γ and object/depth cues, and D_OS is a design parameter with a unimodal trade-off, but these are hyperparameters of a policy, not fitted quantities renamed as predictions of the same data. Self-citations [1], [2] frame the semantic-MAC line and do not supply a uniqueness theorem or force the CARLA ratios. The evaluation’s use of an oracle “perfect view transformation” upper bound is a methodological idealization (and a correctness gap for the end-to-end claim), not circularity: satisfaction under that bound is still a simulated performance metric, not a tautology of the objective. No self-definitional loop, fitted-input-as-prediction, load-bearing self-citation chain, or renamed known identity is present. Score 0 is appropriate.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The central claim rests on simulation modeling choices and several hand-set control knobs rather than a closed-form theorem. Load-bearing domain assumptions include reliable short-range V2V sharing, meaningful FoV overlap in dense traffic, and—most critically—high-quality multi-view reconstruction at the BS. Free parameters (β, γ, η, D_OS, object set, distance/orientation thresholds) directly shape which patches are shared, selected, and prioritized. No new physical entity is postulated; the “invented” pieces are algorithmic constructs (roles, FoV-priority bars, importance score).

free parameters (5)

β, γ in patch importance ω (Eq. 9)
Relative weights of depth and redundancy in scheduling priority; chosen by design, not derived, and they change which patches get airtime.
similarity threshold η
Cutoff for declaring a patch non-redundant after Hungarian matching; directly controls uplink volume.
observation-sharing duration D_OS
Fraction of frame slots reserved for V2V; paper shows unimodal sensitivity and must be estimated by the BS.
preceding/following distance and orientation thresholds (e.g., 50 m, 90°)
Define which pairs share observations; ad hoc geometric pairing rule.
quality threshold Q_th and target-object set
Define “satisfied user” and semantic importance (vehicles, pedestrians, signs, etc.); application-chosen and affect reported ratios.

assumptions (5)

domain assumption Nearby vehicles in dense urban traffic have partially overlapping FoVs that create exploitable spatial redundancy at patch level.
Stated throughout Introduction and §III; without overlap the scheme collapses to ordinary semantic MAC.
domain assumption V2V links can deliver selected patches in a short OS phase without the collision constraints applied to V2I (Eq. 6 relaxed for OS).
§III Observation Sharing Phase; enables local redundancy calculation.
domain assumption GenAI / multi-view methods can transform a neighbor’s patches into the target viewpoint well enough for high-fidelity reconstruction (Eq. 8).
Core of reconstruction model and of the UB evaluation; practical modules not implemented.
domain assumption Each UE transmits on at most one channel per slot and interfering users form known collision cliques (Eqs. 5–6).
Standard MAC hardware/interference model used in scheduling.
ad hoc to paper Hungarian one-to-one matching on SSIM/embedding similarity identifies redundant patches across viewpoints.
§III Data Transmission Phase; algorithmic choice that defines “non-redundant.”

invented entities (2)

FoV-Priority vertical-bar mapping for OS patch order
purpose: Prioritize which patches preceding vehicles share first based on geometric FoV segment overlap.
Paper-specific heuristic (Fig. 1, Algorithm 1); no independent external validation beyond the simulation.
Patch importance score ω combining object coverage, depth, and cross-user reconstruction utility (Eq. 9)
purpose: Centralized BS scheduling priority for non-redundant patches.
Composite score defined for this MAC; parameters free; not a measured physical quantity.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Exploiting Overlapping Fields of View for Redundancy-Aware Uplink Transmission in Vehicular 6G." pith.science (2026). https://pith.science/paper/G6NHKGUG

@misc{pith2026260631715,
  author       = {Pith},
  title        = {Pith review of: Exploiting Overlapping Fields of View for Redundancy-Aware Uplink Transmission in Vehicular 6G},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G6NHKGUG}},
  note         = {Machine review of arXiv:2606.31715}
}

read the original abstract

Emerging uplink-dominant 6G use cases, such as cooperative vehicular streaming, require efficient transmission of high-volume visual data over limited wireless resources. While semantic communications can reduce traffic by prioritizing task-relevant content, most existing approaches treat users independently and therefore overlook spatial redundancy among nearby devices' observations. This paper proposes a semantic-aware multiple access scheme that exploits overlapping fields of view among vehicular users to reduce redundant uplink transmissions. We formulate a joint perception and transmission control problem in which users decide which image patches to transmit, when to transmit them, and over which channel, subject to communication constraints. To address the resulting complexity, we introduce a practical two-phase approach. First, nearby vehicles share selected observation patches over Vehicle-to-Vehicle (V2V) links to calculate inter-user spatial redundancy. Second, users transmit only semantically important, non-redundant patches to the base station, where observations can be reconstructed using the received patches and complementary views from neighboring vehicles. Simulation results in a dense urban vehicular scenario demonstrate that our approach improves the proportion of users who achieve high-fidelity reconstruction, highlighting the potential of semantic-aware multiple access for sustainable and resource-efficient 6G uplink systems.

Figures

Figures reproduced from arXiv: 2606.31715 by the authors.

**Figure 2.** Temporal structure of the system model. Each time frame is aligned [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** High-level overview of the proposed approach pipeline. UEs exploit [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Performance comparison of the centralized semantic-aware state-of [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

18 extracted references

[1]

Semantic Revolution From Communications to Orchestration for 6G: Challenges, Enablers, and Research Directions,

M. Shokrnezhad, H. Mazandarani, T. Taleb, J. Song, and R. Li, “Semantic Revolution From Communications to Orchestration for 6G: Challenges, Enablers, and Research Directions,”IEEE Network, vol. 38, no. 6, pp. 63–71, 2024

2024
[2]

A Semantic-Aware Multiple Access Scheme for Distributed, Dynamic 6G-Based Applica- tions,

H. Mazandarani, M. Shokrnezhad, and T. Taleb, “A Semantic-Aware Multiple Access Scheme for Distributed, Dynamic 6G-Based Applica- tions,” in2024 IEEE Wireless Communications and Networking Con- ference (WCNC), 2024, pp. 1–6

2024
[3]

OFDM-Based Digital Semantic Communication With Importance Awareness,

C. Liu, C. Guo, Y . Yang, W. Ni, and T. Q. S. Quek, “OFDM-Based Digital Semantic Communication With Importance Awareness,”IEEE Transactions on Communications, vol. 72, no. 10, pp. 6301–6315, 2024

2024
[4]

Communicate Less, Synthesize the Rest: Latency-aware Intent-based Generative Semantic Multicasting with Diffusion Models,

X. Liu, M. B. Mashhadi, L. Qiao, Y . Ma, R. Tafazolli, and M. Bennis, “Communicate Less, Synthesize the Rest: Latency-aware Intent-based Generative Semantic Multicasting with Diffusion Models,”IEEE Trans- actions on Vehicular Technology, pp. 1–16, 2026

2026
[5]

Adaptive Semantic Token Communication for Transformer-Based Edge Inference,

A. Devoto, J. Pomponi, M. Merluzzi, P. Di Lorenzo, and S. Scardapane, “Adaptive Semantic Token Communication for Transformer-Based Edge Inference,”IEEE Transactions on Machine Learning in Communications and Networking, vol. 4, pp. 422–437, 2026

2026
[6]

Context-Aware Semantic Communication for the Wireless Networks,

G. Liu, Y . Liu, J. Wang, H. Du, D. Niyato, J. Kang, Z. Xiong, and A. Jamalipour, “Context-Aware Semantic Communication for the Wireless Networks,”IEEE Wireless Communications, vol. 32, no. 5, pp. 20–28, 2025

2025
[7]

Prioritized Information Bottleneck Theoretic Framework With Distributed Online Learning for Edge Video Analytics,

Z. Fang, S. Hu, J. Wang, Y . Deng, X. Chen, and Y . Fang, “Prioritized Information Bottleneck Theoretic Framework With Distributed Online Learning for Edge Video Analytics,”IEEE Transactions on Networking, vol. 33, no. 3, pp. 1203–1219, 2025

2025
[8]

Video TokenCom: Textual Intent-Guided Multi-Rate Video Token Communications with UEP-Based Adaptive Source-Channel Coding,

J. Men, M. B. Mashhadi, N. Wang, Y . Ma, M. Nilsson, and R. Tafazolli, “Video TokenCom: Textual Intent-Guided Multi-Rate Video Token Communications with UEP-Based Adaptive Source-Channel Coding,”

Show all 18 references

[9]

Available: https://arxiv.org/abs/2603.02470

[Online]. Available: https://arxiv.org/abs/2603.02470
[10]

You Only Transmit Once: Unified Generation and Comprehension for Efficient Semantic Communication,

L. Yao, Y . Wei, Y . Zhou, Z. Li, N. Chi, and K.-K. Wong, “You Only Transmit Once: Unified Generation and Comprehension for Efficient Semantic Communication,”IEEE Journal on Selected Areas in Com- munications, vol. 44, pp. 3349–3364, 2026

2026
[11]

Communications Performance Analysis of Wireless Multiple Access Channel With Specially Correlated Sources,

A. Entezami and G. Abed Hodtani, “Communications Performance Analysis of Wireless Multiple Access Channel With Specially Correlated Sources,”IEEE Access, vol. 13, pp. 159 519–159 529, 2025

2025
[12]

Optimizing Visual Sensor Coverage Overlaps for Multiview Surveillance Systems,

A. A. Altahir, V . S. Asirvadam, N. H. B. Hamid, P. Sebastian, N. B. Saad, R. B. Ibrahim, and S. C. Dass, “Optimizing Visual Sensor Coverage Overlaps for Multiview Surveillance Systems,”IEEE Sensors Journal, vol. 18, no. 11, pp. 4544–4552, 2018

2018
[13]

Data-importance-aware power allocation for adaptive real-time communication in computer vision applications,

C. Xu, Y . Ma, R. Tafazolli, and J. Wang, “Data-importance-aware power allocation for adaptive real-time communication in computer vision applications,”IEEE Journal on Selected Areas in Communications, vol. 43, no. 12, pp. 4015–4026, 2025

2025
[14]

At the Dawn of Generative AI Era: A Tutorial-cum-Survey on New Frontiers in 6G Wireless Intelligence,

A. Celik and A. M. Eltawil, “At the Dawn of Generative AI Era: A Tutorial-cum-Survey on New Frontiers in 6G Wireless Intelligence,” IEEE Open Journal of the Communications Society, vol. 5, pp. 2433– 2489, 2024

2024
[15]

Image Segmentation Using Text and Image Prompts,

T. L ¨uddecke and A. Ecker, “Image Segmentation Using Text and Image Prompts,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 7086–7096

2022
[16]

Transformer-based image and video inpainting: current challenges and future directions,

O. Elharrouss, R. Damseh, A. N. Belkacem, E. Badidi, and A. Lakas, “Transformer-based image and video inpainting: current challenges and future directions,” vol. 58, no. 4, p. 124, 2025. [Online]. Available: https://doi.org/10.1007/s10462-024-11075-9

2025 doi
[17]

LoFTR: Detector- Free Local Feature Matching With Transformers,

J. Sun, Z. Shen, Y . Wang, H. Bao, and X. Zhou, “LoFTR: Detector- Free Local Feature Matching With Transformers,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2021, pp. 8922–8931

2021
[18]

Multi-Hop Deep Joint Source- Channel Coding With Deep Hash Distillation for Semantically Aligned Image Recovery,

D. Bergstr ¨om, D. G¨und¨uz, and O. G¨unl¨u, “Multi-Hop Deep Joint Source- Channel Coding With Deep Hash Distillation for Semantically Aligned Image Recovery,” in2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026, pp. 21 321–21 325

2026

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.

[1] [1]

Semantic Revolution From Communications to Orchestration for 6G: Challenges, Enablers, and Research Directions,

M. Shokrnezhad, H. Mazandarani, T. Taleb, J. Song, and R. Li, “Semantic Revolution From Communications to Orchestration for 6G: Challenges, Enablers, and Research Directions,”IEEE Network, vol. 38, no. 6, pp. 63–71, 2024

2024

[2] [2]

A Semantic-Aware Multiple Access Scheme for Distributed, Dynamic 6G-Based Applica- tions,

H. Mazandarani, M. Shokrnezhad, and T. Taleb, “A Semantic-Aware Multiple Access Scheme for Distributed, Dynamic 6G-Based Applica- tions,” in2024 IEEE Wireless Communications and Networking Con- ference (WCNC), 2024, pp. 1–6

2024

[3] [3]

OFDM-Based Digital Semantic Communication With Importance Awareness,

C. Liu, C. Guo, Y . Yang, W. Ni, and T. Q. S. Quek, “OFDM-Based Digital Semantic Communication With Importance Awareness,”IEEE Transactions on Communications, vol. 72, no. 10, pp. 6301–6315, 2024

2024

[4] [4]

Communicate Less, Synthesize the Rest: Latency-aware Intent-based Generative Semantic Multicasting with Diffusion Models,

X. Liu, M. B. Mashhadi, L. Qiao, Y . Ma, R. Tafazolli, and M. Bennis, “Communicate Less, Synthesize the Rest: Latency-aware Intent-based Generative Semantic Multicasting with Diffusion Models,”IEEE Trans- actions on Vehicular Technology, pp. 1–16, 2026

2026

[5] [5]

Adaptive Semantic Token Communication for Transformer-Based Edge Inference,

A. Devoto, J. Pomponi, M. Merluzzi, P. Di Lorenzo, and S. Scardapane, “Adaptive Semantic Token Communication for Transformer-Based Edge Inference,”IEEE Transactions on Machine Learning in Communications and Networking, vol. 4, pp. 422–437, 2026

2026

[6] [6]

Context-Aware Semantic Communication for the Wireless Networks,

G. Liu, Y . Liu, J. Wang, H. Du, D. Niyato, J. Kang, Z. Xiong, and A. Jamalipour, “Context-Aware Semantic Communication for the Wireless Networks,”IEEE Wireless Communications, vol. 32, no. 5, pp. 20–28, 2025

2025

[7] [7]

Prioritized Information Bottleneck Theoretic Framework With Distributed Online Learning for Edge Video Analytics,

Z. Fang, S. Hu, J. Wang, Y . Deng, X. Chen, and Y . Fang, “Prioritized Information Bottleneck Theoretic Framework With Distributed Online Learning for Edge Video Analytics,”IEEE Transactions on Networking, vol. 33, no. 3, pp. 1203–1219, 2025

2025

[8] [8]

Video TokenCom: Textual Intent-Guided Multi-Rate Video Token Communications with UEP-Based Adaptive Source-Channel Coding,

J. Men, M. B. Mashhadi, N. Wang, Y . Ma, M. Nilsson, and R. Tafazolli, “Video TokenCom: Textual Intent-Guided Multi-Rate Video Token Communications with UEP-Based Adaptive Source-Channel Coding,”