Pith. sign in

REVIEW 3 major objections 4 minor 35 references

BANDMAS: Causality-Inspired Semantic Packet Scheduling for Bandwidth-Efficient Multi-Agent Collaboration

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read BANDMAS claims that splitting agent messages into semantic packets and admitting only those with predicted replay value can cut application-layer traffic by 53-77 percent while preserving or improving task utility on three evidence-aggregat

desk verdict BANDMAS is a solid empirical systems paper with a genuinely new packet-level control interface for LLM agent traffic, but its central scheduling mechanism is a heuristic without an optimality or oracle check. read the letter →

arxiv 2608.00458 v1 pith:OG5AOYJC submitted 2026-08-01 cs.MA

Jiangwen Dong , Wanyu Lin This is my paper · ORCID
classification cs.MA
keywords semanticpacketschedulingmulti-agentLLMsystemsbandwidth-efficientcommunicationreplaycausaleffectreceiver-sideadmissioncontrolevidenceaggregationresource-constrainedfrozentrafficevaluation
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

BANDMAS tries to establish that LLM multi-agent communication can be treated as task-oriented traffic. Instead of forwarding complete messages between agents, the system splits each generated message into typed semantic packets - claims, evidence, uncertainty notes, and requests - and forwards only packets whose predicted contribution to the receiver's final output outweighs their resource cost. The paper's central comparison is byte savings versus task utility under an explicit wire cap: with frozen Qwen3-4B producer traffic on SciFact, HotpotQA, and FanOutQA, the causal admission policy reports the highest mean task utility at every selected cap while reducing application-layer bytes by 64.2%, 77.3%, and 53.2% relative to full communication. The paper argues that message length and semantic similarity do not capture task value, whereas a value signal learned from controlled offline replays does. If the claim holds, distributed multi-agent systems can cut communication and receiver-context load substantially without giving up decision quality.

What carries the argument

The load-bearing objects are the semantic packet and the replay causal effect. A semantic packet is an application-layer record carrying one typed segment of a producer message - claim, evidence, uncertainty, or request - plus routing, provenance, and ordering metadata; typed packetization gives the scheduler units fine enough to discard redundant content without tearing a claim away from its support. The replay causal effect labels each packet by half the sufficiency contrast (packet-alone utility minus empty utility) plus half the necessity contrast (full-context utility minus full-context-without-packet utility). These labels are computed only offline, and a frozen shrinkage predictor est

What would settle it

Find or construct a task where two evidence packets each have near-zero RCE individually (so both score low and get dropped) but together shift the receiver from wrong to correct - for example, a HotpotQA question whose answer requires combining one bridge fact from each of two documents. If replay profiling labels both packets near zero, yet the full pair raises task F1 substantially, and BANDMAS's admission drops them while full communication answers correctly, the additivity assumption behind the ranking is violated and the central claim fails.

Watch

Extended reading notes

Core claim

The paper claims that a receiver-side controller can decide which parts of already-generated agent messages cross the communication boundary, and that the right admission signal is a packet's predicted replay causal effect (RCE) rather than its length, topical similarity, or an agent-level pruning decision. RCE is defined per packet as the average of two offline contrasts: sufficiency, the utility gain of the packet alone over an empty context, and necessity, the utility loss of removing the packet from the complete context. BANDMAS computes these labels offline, trains a frozen empirical-Bayes shrinkage predictor from pre-outcome metadata, and then greedily admits complete packets by predic

Load-bearing premise

The scoring assumes each packet's value can be judged independently and added up, so greedily admitting the highest-scoring packets that fit the budget approximates the best subset; the paper explicitly notes this omits higher-order interactions between packets.

Editorial extensions

If this is right

  • Full message forwarding is not the right default: a multi-agent system can cut transmitted evidence traffic by more than half while keeping task utility at or above the full-communication level on the three tested workloads.
  • Admission should be based on predicted contribution to the receiver's output, not on message length or lexical similarity; the paired comparisons favor replay-valued ranking over semantic relevance at matched caps.
  • Typed semantic packets provide better admission granularity than whole messages or sentences, preserving claim-support ties while fitting residual budget.
  • Under queue saturation, adaptive admission with completed-service resource prices reduces p95 latency, but at the expense of task utility; deployments must choose a latency or utility target.
  • Receiver-context tokens drop substantially (roughly 53 to 24 on SciFact, 232 to 68 on HotpotQA), which can lower per-task inference cost even when wall-clock latency gains are modest.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The additive scoring assumption implies complementary packets - each near-zero alone but jointly decisive - will be undervalued and dropped; adding a pair-level or dependency-aware value term is a natural extension the paper does not implement.
  • RCE is defined relative to a fixed receiver and fixed producer traffic, so changing the receiver model would require re-profiling; the Qwen2.5 transfer check changes the producer model, not the receiver.
  • BANDMAS controls content on fixed routes, so it is complementary to agent- or edge-pruning methods; combining route-level pruning with packet-level admission could compound savings, but the paper does not test this.
  • The byte-saving figures count compact application-layer records only; end-to-end network gains depend on transport headers, retransmissions, and encryption, so link-level savings could differ from the reported percentages.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. BANDMAS proposes a receiver-side scheduling layer for LLM multi-agent communication: producer messages are deterministically packetized into typed semantic packets; an offline replay profiler labels each packet with a "replay causal effect" (average of sufficiency and necessity contrasts); a frozen shrinkage predictor estimates this label from pre-outcome metadata; and a scheduler greedily admits packets under wire, context, and deadline constraints. The paper evaluates on frozen Qwen3-4B traffic on SciFact, HotpotQA, and FanOutQA, reporting the highest mean task utility among constrained baselines at selected caps with 53.2–77.3% byte savings, plus a Qwen2.5-7B transfer check, packetization ablations, and adaptive-load experiments.

Significance. If the empirical results are reproducible, BANDMAS offers a promising granular communication-control point between agent/message pruning and token compression, with interpretable packet records and a no-outcome-leakage boundary. The evaluation is careful in several respects: frozen producer traffic, matched caps, exact serialized-batch accounting, disjoint profile/validation/test partitions, forbidden gold features in the predictor, paired tests with Holm correction, and explicit limitation statements. The main weakness is that the paper's central scheduling mechanism—greedy admission by predicted per-packet RCE—is not shown to approximate the stated set-level constrained objective, leaving a structural gap between the formalism and the algorithm.

major comments (3)
  1. [III–IV (Eqs. (5)–(6), Algorithm 1)] The set-level objective (5) is replaced in Algorithm 1 by ranking packets with independent scores s(p) and greedily admitting until the wire cap. RCE in Eq. (6) is a two-point contrast, and the paper explicitly states it omits higher-order interactions and is "a bounded ranking target rather than an additive decomposition". For nonlinear set utility, per-packet values can misorder complementary packets, and greedy-by-value has no approximation guarantee even for a knapsack. No oracle or near-oracle comparison (e.g., exhaustive/DP optimization on profile/validation tasks) is reported, so the causal scheduling mechanism is not validated. Please add such a comparison or substantially weaken the causal-scheduling claim.
  2. [V.B/Table I] The headline claim "highest mean task metric among constrained methods" is based on only three split seeds and no paired significance tests against AgentPrune/SafeSieve; Table I intervals overlap on FanOutQA (causal 0.234±0.047 vs SafeSieve 0.209±0.013), and Table II tests only causal vs semantic relevance. Although V.B acknowledges the insufficiency, the abstract states the result without qualification. Please add paired tests/confidence intervals against the strongest external baselines or restrict the claim to descriptive ranking.
  3. [IV.C/Table III] The frozen predictor's rank correlation (0.480–0.538) and useful-packet recall (0.522–0.582) are moderate. Because RCE labels are defined in Full/Empty contexts, they also differ from the deployed subset distribution. This is not fatal by itself, but it reinforces the need for subset-level validation: the final selected subsets should be compared with oracle-optimal subsets on held-out tasks to show that the moderate predictor plus greedy admission yields near-optimal utility.
minor comments (4)
  1. [I/IV.D] The contribution bullet "exact wire-cap admission" is misleading: Algorithm 1 is greedy, and only the serialized-batch feasibility check is exact. Please rephrase to "exact serialized-batch feasibility checks".
  2. [IV.C] The hybrid weights λ1=0.75, λ2=0.25 in Eq. (8) are fixed without sensitivity analysis; since hybrid is close to causal on several rows, a small robustness check would be useful.
  3. [V.B (Fig. 3)] First-order error propagation with zero covariance is a heuristic; the caption states this, but please make the limitation more prominent or replace with bootstrap confidence intervals.
  4. [V.C (Table III)] Please define "useful-packet recall" precisely (presumably recall of positive-RCE packets at a chosen threshold) so the diagnostic is reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: RCE predictor is trained on disjoint profile/validation partitions and evaluated on frozen final-test tasks, with gold outcomes explicitly forbidden.

full rationale

The paper's central claim is an empirical comparison under frozen producer traffic. The replay causal effect (RCE) in Eq. (6) is defined in terms of the same utility U used in evaluation, but the predictor is trained only on profile/validation tasks and then frozen; final-test outcomes, labels, and RCE values are explicitly forbidden features (Sec. IV.C, V.A). The online scheduler uses only pre-outcome metadata, so the final-test task utility is a genuine generalization measure, not a re-statement of the training objective. The greedy admission by predicted RCE is an approximation to the set-level objective Eq. (5), and the paper acknowledges that RCE omits arbitrary higher-order interactions (Sec. III, IV.C); this is a correctness/optimality limitation, not circularity. Self-citations, such as [30] for counterfactual removal, are motivational and not load-bearing: Eq. (6) defines the contrasts directly. No equation reduces to its own inputs by construction, and the evaluation is self-contained against external benchmarks. Therefore the circularity score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 1 invented entities

The key uncharged assumptions are the additive use of per-packet values, transferability of replay labels, equal task and evidence weighting, completeness of the four-field message contract, and sufficiency of the metadata feature hierarchy.

free parameters (3)
  • alpha shrinkage strength = 6
    Eq (7): fixed across all benchmarks; hand-set rather than derived or tuned per task.
  • hybrid weights lambda_1, lambda_2 = 0.75, 0.25
    Eq (8): reported configuration; no selection procedure or sensitivity analysis is given.
  • adaptive price learning rates eta_k = unspecified
    Eq (9): step sizes for resource price updates are not reported in the manuscript.
assumptions (5)
  • ad hoc to paper Set-level utility U can be decomposed into additive per-packet values; greedy admission by predicted RCE approximates constrained maximization of U(F_x(S)).
    Sec IV.C: RCE is defined per packet and omits arbitrary higher-order interactions; Algorithm 1 then ranks and admits by predicted RCE. This is a structural assumption not proved.
  • domain assumption Replay labels computed on profile/validation partitions with frozen traffic generalize to final-test tasks.
    Sec IV.C and V.A: the predictor is frozen before final-test scheduling; no distribution-matching argument is provided.
  • domain assumption Equal weighting of task and evidence utility in Eq (4) is a valid supervisory objective.
    Eq (4): U = 0.5 T + 0.5 E. The weighting is arbitrary but applied consistently to all policies.
  • domain assumption The four structured fields (Claim, Evidence, Uncertainty, Request) cover the information needed for downstream decisions.
    Sec IV.B: readers must emit these fields under a prompt contract; if the contract omits relevant content, packetization cannot recover it.
  • domain assumption The frozen-scorer feature hierarchy with empirical Bayes shrinkage (Eq 7) is sufficient for pre-outcome packet value estimation.
    Sec IV.C: features are packet type, route, lexical markers, size, and turn position; reported rank correlations of 0.48-0.54 indicate incomplete signal.
invented entities (1)
  • semantic packet independent evidence
    purpose: Application-layer schedulable unit carrying one message segment plus routing and reconstruction metadata
    It is a concrete, implemented, serialized record; independent implementations can parse and verify it, and the code link is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BANDMAS: Causality-Inspired Semantic Packet Scheduling for Bandwidth-Efficient Multi-Agent Collaboration." pith.science (2026). https://pith.science/paper/OG5AOYJC

@misc{pith2026260800458,
  author       = {Pith},
  title        = {Pith review of: BANDMAS: Causality-Inspired Semantic Packet Scheduling for Bandwidth-Efficient Multi-Agent Collaboration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OG5AOYJC}},
  note         = {Machine review of arXiv:2608.00458}
}
read the original abstract

LLM-based multi-agent systems make decisions based on the aggregated information via exchanging messages across specialized agents. Forwarding every generated message among agents increases application-layer traffic. Yet, it introduces tremendous input tokens for agent processing, potentially raising inference latency and computational overhead. Existing approaches attempt to address the above issues by pruning agents or discarding redundant messages. Nevertheless, such agent-level or message-level optimization results in insufficient evidence supporting for final decisions or still containing redundant message transmissions. To address these challenges, we propose BANDMAS, a multi-agent collaboration framework that models inter-agent communications as task-oriented traffic, which enables efficient transmission via causality-inspired replay valuation. Specifically, we decompose messages into several data packets by analyzing their semantic features such as evidence and requests. The system only transmits these packets if their predicted replay-derived contribution exceeds their resource cost. Consequently, BANDMAS is able to adaptively schedule communication packets while adhering to bandwidth, latency, deadline, and receiver context constraints. On frozen Qwen3-4B traffic across SciFact, HotpotQA, and FanOutQA, our framework reduces application-layer bytes by 53.2\% to 77.3\% at selected caps and attains the highest mean task metric among constrained methods on all three workloads.

Figures

Figures reproduced from arXiv: 2608.00458 by the authors.

Figure 1
Figure 1. Communication-control boundary. Message-level workflows forward [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. BANDMAS online framework. Generated messages become typed [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Task utility per transmitted application-layer KiB ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Packetization utility–wire frontiers under identical messages and lexical selection. Points denote budgets 0.25, 0.50, 0.75, and Full; bars show one [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Task utility and p95 end-to-end latency over normalized offered load. Curves aggregate three split seeds; bands are 95% hierarchical-bootstrap intervals. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 33 canonical work pages

  1. [1]

    Autogen: Enabling next-gen LLM applications via multi- agent conversations,

    Q. Wu, G. Bansal, J. Zhang, Y . Wu, B. Li, E. Zhu, L. Jiang, X. Zhang, S. Zhang, J. Liu, A. H. Awadallah, R. W. White, D. Burger, and C. Wang, “Autogen: Enabling next-gen LLM applications via multi- agent conversations,” inFirst Conference on Language Modeling, 2024

    work page 2024

  2. [2]

    MetaGPT: Meta programming for a multi-agent collaborative framework,

    S. Hong, M. Zhuge, J. Chen, X. Zheng, Y . Cheng, J. Wang, C. Zhang, Z. Wang, S. K. S. Yau, Z. Lin, L. Zhou, C. Ran, L. Xiao, C. Wu, and J. Schmidhuber, “MetaGPT: Meta programming for a multi-agent collaborative framework,” inInternational Conference on Learning Representations, 2024

    work page 2024

  3. [3]

    Cut the crap: An economical communication pipeline for LLM-based multi-agent systems,

    G. Zhang, Y . Yue, Z. Li, S. Yun, G. Wan, K. Wang, D. Cheng, J. X. Yu, and T. Chen, “Cut the crap: An economical communication pipeline for LLM-based multi-agent systems,” inThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  4. [4]

    AgentDropout: Dynamic agent elimination for token-efficient and high- performance LLM-based multi-agent collaboration,

    Z. Wang, Y . Wang, X. Liu, L. Ding, M. Zhang, J. Liu, and M. Zhang, “AgentDropout: Dynamic agent elimination for token-efficient and high- performance LLM-based multi-agent collaboration,” inProceedings of the 63rd Annual Meeting of the Association for Computational Linguis- tics (Volume 1: Long Papers), 2025, pp. 24 013–24 035

    work page 2025

  5. [5]

    SafeSieve: From heuristics to experience in progressive pruning for LLM-based multi-agent communication,

    R. Zhang, X. Zhao, R. Wang, S. Chen, G. Zhang, A. Zhang, K. Wang, and Q. Wen, “SafeSieve: From heuristics to experience in progressive pruning for LLM-based multi-agent communication,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 35, pp. 29 892– 29 900, 2026

    work page 2026

  6. [6]

    Improving multi-agent debate with sparse communication topology,

    Y . Li, Y . Du, J. Zhang, L. Hou, P. Grabowski, Y . Li, and E. Ie, “Improving multi-agent debate with sparse communication topology,” inFindings of the Association for Computational Linguistics: EMNLP 2024, 2024, pp. 7281–7294

    work page 2024

  7. [7]

    G-Designer: Architecting multi-agent communication topologies via graph neural networks,

    G. Zhang, Y . Yue, X. Sun, G. Wan, M. Yu, J. Fang, K. Wang, T. Chen, and D. Cheng, “G-Designer: Architecting multi-agent communication topologies via graph neural networks,” inProceedings of the 42nd International Conference on Machine Learning, vol. 267. PMLR, 2025, pp. 76 678–76 692

    work page 2025

  8. [8]

    MasRouter: Learning to route LLMs for multi-agent systems,

    Y . Yue, G. Zhang, B. Liu, G. Wan, K. Wang, D. Cheng, and Y . Qi, “MasRouter: Learning to route LLMs for multi-agent systems,” in Proceedings of the 63rd Annual Meeting of the Association for Compu- tational Linguistics (Volume 1: Long Papers), 2025, pp. 15 549–15 572

    work page 2025

Show all 35 references
  1. [9]

    LLMLingua: Com- pressing prompts for accelerated inference of large language models,

    H. Jiang, Q. Wu, C.-Y . Lin, Y . Yang, and L. Qiu, “LLMLingua: Com- pressing prompts for accelerated inference of large language models,” in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023, pp. 13 358–13 376

    2023

  2. [10]

    LongLLMLingua: Accelerating and enhancing LLMs in long context scenarios via prompt compression,

    H. Jiang, Q. Wu, X. Luo, D. Li, C.-Y . Lin, Y . Yang, and L. Qiu, “LongLLMLingua: Accelerating and enhancing LLMs in long context scenarios via prompt compression,” inProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers...

    2024

  3. [11]

    LLMLingua-2: Data distillation for efficient and faithful task-agnostic prompt compres- sion,

    Z. Pan, Q. Wu, H. Jiang, M. Xia, X. Luo, J. Zhang, Q. Lin, V . R ¨uhle, Y . Yang, C.-Y . Lin, H. V . Zhao, L. Qiu, and D. Zhang, “LLMLingua-2: Data distillation for efficient and faithful task-agnostic prompt compres- sion,” inFindings of the Association for Computational Ling...

    2024

  4. [12]

    KVCOMM: Online cross-context KV- cache communication for efficient LLM-based multi-agent systems,

    H. Ye, Z. Gao, M. Ma, Q. Wang, Y . Fu, M.-Y . Chung, Y . Lin, Z. Liu, J. Zhang, D. Zhuo, and Y . Chen, “KVCOMM: Online cross-context KV- cache communication for efficient LLM-based multi-agent systems,” in Advances in Neural Information Processing Systems, vol. 38, 2025

    2025

  5. [13]

    Q-KVComm: Efficient multi-agent communication via adaptive KV cache compression,

    B. Kriuk and L. Ng, “Q-KVComm: Efficient multi-agent communication via adaptive KV cache compression,” in2026 Second International Conference on Multi-Agent Systems for Collaborative Intelligence, 2026, pp. 15–21

    2026

  6. [14]

    Evolving idea graphs with learnable edits-and-commits for multi-agent scientific ideation,

    J. Dong, B. Li, and W. Lin, “Evolving idea graphs with learnable edits-and-commits for multi-agent scientific ideation,”arXiv preprint arXiv:2605.04922, 2026

    2026 arXiv

  7. [15]

    Multi-agent architecture search via agentic supernet,

    G. Zhang, L. Niu, J. Fang, K. Wang, L. Bai, and X. Wang, “Multi-agent architecture search via agentic supernet,” inProceedings of the 42nd International Conference on Machine Learning, vol. 267. PMLR, 2025, pp. 75 834–75 852

    2025

  8. [16]

    S-DAG: A subject-based directed acyclic graph for multi-agent heterogeneous reasoning,

    J. Dong, Z. Lin, W. Lin, and M. Zhang, “S-DAG: A subject-based directed acyclic graph for multi-agent heterogeneous reasoning,”Pro- ceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 35, pp. 29 394–29 402, 2026

    2026

  9. [17]

    Learning multiagent com- munication with backpropagation,

    S. Sukhbaatar, A. Szlam, and R. Fergus, “Learning multiagent com- munication with backpropagation,” inAdvances in Neural Information Processing Systems, vol. 29, 2016

    2016

  10. [18]

    Learning to communicate with deep multi-agent reinforcement learning,

    J. N. Foerster, Y . M. Assael, N. de Freitas, and S. Whiteson, “Learning to communicate with deep multi-agent reinforcement learning,” in Advances in Neural Information Processing Systems, vol. 29, 2016

    2016

  11. [19]

    TarMAC: Targeted multi-agent communication,

    A. Das, T. Gervet, J. Romoff, D. Batra, D. Parikh, M. Rabbat, and J. Pineau, “TarMAC: Targeted multi-agent communication,” inProceed- ings of the 36th International Conference on Machine Learning, vol. 97. PMLR, 2019, pp. 1538–1546

    2019

  12. [20]

    Learning when to communicate at scale in multiagent cooperative and competitive tasks,

    A. Singh, T. Jain, and S. Sukhbaatar, “Learning when to communicate at scale in multiagent cooperative and competitive tasks,” inInternational Conference on Learning Representations, 2019

    2019

  13. [21]

    Learning to schedule communication in multi- agent reinforcement learning,

    D. Kim, S.-W. Moon, D. E. Hostallero, W. J. Kang, S.-H. Lee, K. Son, and Y . Yi, “Learning to schedule communication in multi- agent reinforcement learning,” inInternational Conference on Learning Representations, 2019

    2019

  14. [22]

    Deep learning enabled semantic communication systems,

    H. Xie, Z. Qin, G. Y . Li, and B.-H. Juang, “Deep learning enabled semantic communication systems,”IEEE Transactions on Signal Pro- cessing, vol. 69, pp. 2663–2675, 2021

    2021

  15. [23]

    Learning-based scheduling for information gathering with QoS constraints,

    Q. Liu, W. Xu, and Z. Fang, “Learning-based scheduling for information gathering with QoS constraints,” inIEEE INFOCOM 2024 - IEEE Conference on Computer Communications, 2024, pp. 431–440

    2024

  16. [24]

    AdaSem: Adaptive goal-oriented semantic communications for end-to-end camera relocalization,

    Q. Liao and T.-Y . Tung, “AdaSem: Adaptive goal-oriented semantic communications for end-to-end camera relocalization,” inIEEE INFO- COM 2024 - IEEE Conference on Computer Communications, 2024, pp. 1111–1120

    2024

  17. [25]

    HALO: Semantic- aware distributed LLM inference in lossy edge network,

    P. Zheng, W. Xu, H. Wang, J. Chen, and X. Shen, “HALO: Semantic- aware distributed LLM inference in lossy edge network,” inIEEE INFOCOM 2026 - IEEE Conference on Computer Communications, 2026, pp. 1–10

    2026

  18. [26]

    Semantic packet aggregation for token communication via genetic beam search,

    S. Lee, J. Park, J. Choi, and H. Park, “Semantic packet aggregation for token communication via genetic beam search,” in2025 IEEE 26th International Workshop on Signal Processing and Artificial Intelligence for Wireless Communications (SPAWC), 2025, pp. 1–5

    2025

  19. [27]

    Distributed inference with deep learning models across heterogeneous edge devices,

    C. Hu and B. Li, “Distributed inference with deep learning models across heterogeneous edge devices,” inIEEE INFOCOM 2022 - IEEE Conference on Computer Communications, 2022, pp. 330–339

    2022

  20. [28]

    Titanic: Towards production federated learning with large language models,

    N. Su, C. Hu, B. Li, and B. Li, “Titanic: Towards production federated learning with large language models,” inIEEE INFOCOM 2024 - IEEE Conference on Computer Communications, 2024, pp. 611–620

    2024

  21. [29]

    Hybridflow: Resource-adaptive subtask routing for efficient edge-cloud LLM inference,

    J. Dong, J. Li, T. Zheng, and W. Lin, “Hybridflow: Resource-adaptive subtask routing for efficient edge-cloud LLM inference,” inForty-third International Conference on Machine Learning, 2026

    2026

  22. [30]

    Generative causal explanations for graph neural networks,

    W. Lin, H. Lan, and B. Li, “Generative causal explanations for graph neural networks,” inProceedings of the 38th International Conference on Machine Learning, vol. 139. PMLR, 2021, pp. 6666–6679

    2021

  23. [31]

    Fact or fiction: Verifying scientific claims,

    D. Wadden, S. Lin, K. Lo, L. L. Wang, M. van Zuylen, A. Cohan, and H. Hajishirzi, “Fact or fiction: Verifying scientific claims,” in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, 2020, pp. 7534–7550

    2020

  24. [32]

    HotpotQA: A dataset for diverse, explainable multi- hop question answering,

    Z. Yang, P. Qi, S. Zhang, Y . Bengio, W. W. Cohen, R. Salakhutdinov, and C. D. Manning, “HotpotQA: A dataset for diverse, explainable multi- hop question answering,” inProceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, 2018, pp. 2369– 2380

    2018

  25. [33]

    FanOutQA: A multi-hop, multi-document question answering benchmark for large language models,

    A. Zhu, A. Hwang, L. Dugan, and C. Callison-Burch, “FanOutQA: A multi-hop, multi-document question answering benchmark for large language models,” inProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), 2024, pp. 18–37

    2024

  26. [34]

    Poisoning attacks on deep learning based wireless traffic prediction,

    T. Zheng and B. Li, “Poisoning attacks on deep learning based wireless traffic prediction,” inIEEE INFOCOM 2022 - IEEE Conference on Computer Communications, 2022, pp. 660–669

    2022

  27. [35]

    Medley: Predicting social trust in time-varying online social networks,

    W. Lin and B. Li, “Medley: Predicting social trust in time-varying online social networks,” inIEEE INFOCOM 2021 - IEEE Conference on Computer Communications, 2021, pp. 1–10

    2021

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.