Pith. sign in

REVIEW 6 major objections 9 minor 1 cited by

Scene Understanding Enabled Semantic Communication with Open Channel Coding

T0 review · 6 major / 9 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A scene graph transmitted as text tokens over QAM symbols and repaired by an LLM at the receiver can answer visual questions with roughly 4,160 symbols instead of the 16 million needed by image compression.

desk verdict Plausible scene-graph transmission system with a dramatic efficiency claim that the current evaluation doesn't yet back up. read the letter →

arxiv 2501.14520 v1 pith:AHGCAGJJ submitted 2025-01-24 eess.SP cs.CV

Zhe Xiang , Fei Yu , Quan Deng , Yuandi Li , Zhiguo Wan This is my paper · ORCID
classification eess.SPcs.CV
keywords semanticcommunicationscenegraphencodingopenchannelcodingvisualquestionansweringlargelanguagemodelsretrieval-augmentedgenerationQAMmodulation6Gnetworks
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

The paper proposes OpenSC, a semantic communication system that turns an image into a scene graph and transmits that graph's text rather than the image pixels. On the transmit side, scene graph triples are tokenized with a pretrained transformer tokenizer, mapped to token IDs, and converted to QAM constellation symbols; no learned channel code is trained. On the receive side, the symbols are demodulated back to tokens, and a large language model with retrieval-augmented generation repairs the corrupted text and answers visual questions. The paper reports that on a 400-image aerial urban dataset, this pipeline answers category, quantity, location, and relationship questions with higher recall and F1 than conventional image compression and prior semantic communication baselines, while sending about 4,160 symbols per item instead of the 16,242,744 symbols used by JPEG with LDPC coding. If the claim holds, visual semantic communication can shift from fixed task-specific codebooks to shared language-model knowledge.

What carries the argument

The load-bearing object is the scene-graph-to-text pipeline used as an open channel code. An image is first parsed into subject-predicate-object triples by a prototype-based scene graph generator; the triples are tokenized into subword units and mapped through the token vocabulary into binary strings, and those strings are segmented into $m$-bit chunks that become QAM constellation symbols. The receiver estimates the channel (zero-forcing linear MMSE with channel state information, or minimum-Euclidean-distance demodulation without it), converts symbols back to token IDs, and reconstructs the text. The paper's named innovation, open channel coding, is the replacement of a trained joint source-channel code with this token-to-constellation mapping plus receiver-side semantic repair: the LLM is prompted with the reconstructed text augmented by retrieved chunks from a vector database built from the received scene graph information, and it outputs a JSON summary (number, location, relationship) that supports final question answering. This is what lets the system claim knowledge-agnostic, training-free channel coding.

What would settle it

Disable the receiver-side LLM and retrieval-augmented repair, feed the demodulated token stream directly into a text-only question answering model, and measure recall on the four VQA question types; if recall collapses toward the ablation levels reported when the open coder is removed, then the LLM, not the open channel coding, is what makes the system work.

Watch

Extended reading notes

Core claim

OpenSC's central claim is that a scene graph, written as subject-predicate-object triples, is a sufficient semantic representation for visual question answering over wireless channels, and that the text itself can be protected not by error-correcting codes but by a large language model at the receiver. The transmitter converts the graph text into token IDs and then into QAM symbols; the receiver reverses this mapping and feeds the reconstructed text, together with structured JSON fields for number, location, and relationship, into a retrieval-augmented large language model that fills gaps and answers the question without ever seeing the original image. In the reported experiments on the aerial urban dataset, this yields recall values of 0.6401 for category, 0.3255 for quantity, 0.3227 for location, and 0.1398 for relationship at 18 dB SNR with the Qwen-plus receiver model, and it outperforms the baselines across most SNR points in both AWGN and Rayleigh channels. The paper also reports that changing the modulation scheme (BPSK, 4QAM, 16QAM) barely changes performance, and that swapping the receiver LLM causes only minor variation, which it reads as evidence that the semantic code, not the channel code or the answer model, carries the robustness.

Load-bearing premise

The entire efficiency gain rests on the premise that the receiver's large language model can repair a corrupted scene graph text well enough to answer visual questions, even though the symbols carry no error-correcting code and the original image is never transmitted.

Editorial extensions

If this is right

  • If correct, the transmitter does not need to send image pixels at all for visual question answering: the scene graph text carries the task-relevant semantics, which implies large bandwidth savings for visual tasks on limited-resource links.
  • Because the channel coding stage has no learned parameters, the same OpenSC transmitter can in principle serve different downstream tasks and different receiver LLMs without retraining, as long as the scene graph vocabulary covers the task.
  • The near-invariance to modulation order means the system can trade spectral efficiency for robustness by choosing BPSK, 4QAM, or 16QAM after measuring the channel, rather than fixing one codebook.
  • The reported symbol count implies that structured text semantics can be more bandwidth-efficient than compressed image transmission by several orders of magnitude in this aerial-image setting.

Reading between the lines

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

  • Editorial inference: the 4,160-symbol figure is meaningful only relative to the scene graph's token count; for images with richer scenes or questions about attributes outside the number/location/relationship JSON slots, the symbol budget and the efficiency ratio would both grow.
  • Editorial inference: a direct stress test would rebuild the retrieval vector database from the corrupted received text instead of from clean scene graph chunks; if performance drops sharply, the reported robustness is partly an artifact of clean-side retrieval.
  • Editorial inference: the approach transfers to other visual domains only when a scene graph generator with matching object and predicate categories is available; domain shift in the graph vocabulary would silently drop the semantics needed by downstream questions.
  • Editorial inference: a natural extension is to compare OpenSC against a learned joint source-channel code that transmits the same scene graph text, isolating whether the LLM's repair, rather than the scene graph format itself, causes the accuracy gain.
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.

Referee Report

6 major / 9 minor

Summary. The paper proposes OpenSC, a semantic communication system for visual question answering (VQA). At the transmitter, a scene graph is extracted from the input image, the graph text is tokenized with WordPiece, and the BERT token IDs are mapped directly to QAM symbols without forward error correction (Eq. (11)). The receiver demodulates the symbols, reconstructs token IDs and text, and then uses an LLM with retrieval-augmented generation to answer questions about category, quantity, location, and relationship. The RAG database is built from the received chunks themselves. Experiments on the AUG dataset compare OpenSC with DeepSC, MJCMSC, LLM-SC, and traditional schemes (5-bit+RS, Huffman+RS, JPEG+LDPC) over AWGN and Rayleigh channels, reporting recall/F1 curves and symbol counts. The headline efficiency result is that OpenSC transmits 4,160 symbols per item versus 16,242,744 for JPEG-LDPC while achieving higher recall and F1.

Significance. If substantiated, the paper would show a practically important result: a scene-graph description of an image can be communicated with orders-of-magnitude fewer symbols than JPEG-compressed image transmission while supporting VQA more accurately. The integration of scene graph generation, BERT tokenization, QAM modulation, and LLM-based semantic decoding is a plausible and timely system design. The paper also contains useful ablations over LLM choices and comparisons against several baselines, and it is clearly written in its high-level architecture. However, the central robustness and efficiency claims currently rest on an unspecified evaluation protocol, a self-referential RAG mechanism, and the absence of any channel-error measurement. The paper does not provide code, data, or machine-checked proofs, so independent verification is not possible from the manuscript alone.

major comments (6)
  1. [Sec. 4.2, Eq. (11), and Figs. 5–6] The transmitter maps BERT token IDs directly to QAM symbols with no forward error correction, so a single bit error changes a WordPiece token and there is no channel code to recover it. The paper nevertheless claims robustness across AWGN and Rayleigh channels, but it reports no token error rate or bit error rate and no noiseless scene-graph upper bound. Moreover, Figs. 5–6 show nearly identical recall/F1 for BPSK, 4QAM, and 16QAM across SNR, which is unexplained if symbol errors are actually present at the tested SNRs; at a fixed symbol SNR these modulations have very different symbol error rates. Please report token error rates, add a noiseless upper bound, and explain the modulation invariance (or show that the SNR range is effectively error-free).
  2. [Sec. 4.3.5] The RAG module builds its vector database from chunks of the received information itself and retrieves the four closest chunks to seed the LLM. Because the database is derived from the same corrupted stream, it provides no independent evidence about the original image; if a transmitted attribute is corrupted to another plausible word, retrieval will reinforce the wrong value. The claimed 'semantic enhancement' is therefore not separable from the LLM's language priors. Please evaluate the system with a clean knowledge base built from ground-truth scene graphs, and also with retrieval disabled, to quantify what the RAG step actually contributes.
  3. [Sec. 5.1.2 and Sec. 5.3] The evaluation protocol for recall and F1 is not specified: the paper does not state the test-set size, the number of questions per type, or the matching procedure between open-ended LLM answers and ground-truth objects, quantities, locations, and relationships. Figures 3–8 report point estimates only, with no error bars, confidence intervals, or significance tests. The word 'significant' in the abstract and conclusion is therefore not statistically supported. Please describe the exact evaluation protocol and report variability measures.
  4. [Sec. 3 vs. Sec. 4.2 and Algorithm 1] The system model in Eqs. (2)–(3) describes trainable channel encoder and decoder parameters γ and θ, but Section 4.2 and Algorithm 1 state that channel encoding and decoding use a fixed LLM with constellation tokens and require no additional training. This mismatch means that what the paper calls 'open channel coding' is in fact uncoded QAM modulation plus LLM-based post-processing, not a trainable channel code. Please either instantiate the trainable channel code described in the system model or revise the system model to reflect the actual uncoded scheme.
  5. [Sec. 5.5 and Table 3] The headline efficiency ratio is misstated. 4,160 / 16,242,744 ≈ 2.56 × 10^-4 = 0.0256%, not '2.6 × 10^-4%' as claimed in the text, and Table 3 labels the column 'Ratio (%)' but enters '2.6 × 10^-4' as if it were a dimensionless ratio. The qualitative conclusion is unchanged, but the numbers should be corrected. In addition, please state the SNR/channel conditions assumed for the symbol-count comparison and confirm that the baselines achieve comparable VQA performance at that operating point.
  6. [Sec. 5.4.1 and Table 2] The ablation rows in Table 2 are hard to interpret: the first row has SSC absent but OCC present, even though OCC is described as transmitting scene-graph tokens, and the third row reports no metrics at all. Please define what each ablation actually transmits and reports, and provide complete results for all rows, including the missing values.
minor comments (9)
  1. [Sec. 4.2 title] The section title contains a typo: 'Open Channel Condings' should be 'Open Channel Coding'.
  2. [Algorithm 1 and Sec. 5.1.3] Algorithm 1 has 'taining' instead of 'training', and Section 5.1.3 contains the incomplete sentence 'Learning rate decay occurs sets the 90,000 and 120,000 iterations.'
  3. [Table 1] The column label 'BLUE (1-gram)' should be 'BLEU (1-gram)'.
  4. [Figure captions] Figures 3–8 refer to 'A WGN' with a space; this should be 'AWGN'.
  5. [Sec. 5.5 and Table 3 caption] The paper alternates between 'LDPC' and 'LPDC'; the latter is a misspelling that should be corrected throughout.
  6. [Sec. 3.2.2] Equation (5) is a verbatim repeat of Eq. (4); one of them should be removed or the two should be differentiated.
  7. [Title page and metadata] The ACM template metadata still contains the placeholder conference name, 2018 date, and DOI; this should be updated before submission.
  8. [Sec. 2.1] The sentence 'Similarly, proposed a semantic communication framework (LAM-SC)...' is missing the author names before 'proposed'.
  9. [Sec. 5.7] The visualization section claims the proposed method produces 'only three minor errors', but no protocol for counting errors is given.

Circularity Check

1 steps flagged · score 6.0 of 10

RAG-based semantic enhancement at the receiver is self-referential: the database is built from the received stream itself, so claimed error correction cannot supply lost information.

  1. self definitional [Section 4.3.2 and Section 4.3.5 (Eqs. 14-19, vector database construction)]
    "The reconstructed information may be incomplete or contain errors, so we supplement it with previously structured knowledge. This structured information typically includes object attributes such as quantity, location, and relationships, and is organized in the form of a Scene Graph. ... Organize objects, relationships, and attributes into chunks representing specific categories with their quantity, location, and relationships. Embed these chunks into a high-dimensional vector space to get dense vector representations. Store these vectors in a temporary database, linking each to its chunk."

    The receiver's only input is the demodulated symbol stream (Eq. 14 reconstructs tokens from it), and the final_information JSON (Eq. 19) is derived from that same stream. The 'previously structured knowledge' and the RAG chunks are therefore both functions of the received, possibly corrupted content. There is no original image, no clean side information, and no external knowledge base at the receiver. The LLM is asked to 'correct and enrich' a corrupted message using a vector database populated from that same corrupted message. Any retrieved chunk is a reshuffling of already-received content, so the semantic-enhancement step cannot restore information lost in the channel.

full rationale

The efficiency comparison (4,160 symbols vs. 16,242,744 for JPEG-LDPC) is not circular: it compares a scene-graph token stream against a compressed image, and the scene-graph encoder is trained on AUG data independently of the VQA answers. The no-FEC QAM mapping is a design choice, not a circularity. However, the paper's central robustness/error-correction story is partially circular. Section 4.3.5 builds the retrieval database from the received information itself, so the LLM-RAG enhancement is self-referential by construction and cannot add information beyond what was already received. The paper also reports no token error rates or noiseless upper bounds, leaving the near-identical BPSK/4QAM/16QAM results unexplained; that absence is a correctness/evidence gap rather than a separate circular step. Overall, the efficiency claim stands, but the robustness claim is validated only through a self-referential loop, giving a partial circularity score of 6.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The system rests on learned components (scene graph generator, BERT, LLM) whose weights are taken from prior work, and on a hand-picked retrieval count and modulation width. The central claim is an empirical system demonstration, not a law; free parameters are the experimental knobs that shape the reported recall values.

free parameters (3)
  • Number of chunks retrieved in RAG = 4
    Section 4.3.5 fixes the nearest-neighbor count to four; no ablation over this value is reported, and it directly influences the prompt given to the VQALLM.
  • Temperature tau in contrastive loss = learnable
    Equation (8) introduces a learnable tau for the prototype-guided loss; its value after training is not reported, yet it shapes the learned semantic space.
  • Modulation order and bit length m = BPSK, 4QAM, 16QAM
    Section 4.2 maps m-bit token IDs to QAM symbols; m is not specified, and the experiments shift modulation schemes without specifying how m changes. This is a hand-chosen system parameter.
assumptions (4)
  • domain assumption Pre-trained BERT tokenizer and model provide a universal, open vocabulary that both transmitter and receiver share.
    Section 4.2 relies on WordPiece tokenization and BERT token IDs to map scene graph text to symbols; if the vocabulary is not shared or the token IDs are not stable, the mapping breaks.
  • domain assumption The AUG dataset's ground truth scene graphs are complete and accurate enough to serve as the reference for recall and F1.
    Section 5.1.1 describes AUG with 77 object categories and 63 relationship types; the evaluation compares model answers against these annotations, so errors in the annotations directly corrupt the reported scores.
  • domain assumption The LLM used at the receiver (Qwen-plus) answers VQA questions consistently and without using information outside the provided prompt.
    Section 5.2 states Qwen-plus is used for all instances; the paper provides no control for the LLM's internal knowledge or nondeterminism, yet the VQA accuracy depends on it.
  • domain assumption AWGN and Rayleigh channel models used in the simulations accurately represent the target deployment environment.
    Section 3.2.1 defines Y = H X + N with Gaussian noise; the paper's entire experimental section evaluates only these models, not real wireless channels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scene Understanding Enabled Semantic Communication with Open Channel Coding." pith.science (2026). https://pith.science/paper/AHGCAGJJ

@misc{pith2026250114520,
  author       = {Pith},
  title        = {Pith review of: Scene Understanding Enabled Semantic Communication with Open Channel Coding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AHGCAGJJ}},
  note         = {Machine review of arXiv:2501.14520}
}
read the original abstract

As communication systems transition from symbol transmission to conveying meaningful information, sixth-generation (6G) networks emphasize semantic communication. This approach prioritizes high-level semantic information, improving robustness and reducing redundancy across modalities like text, speech, and images. However, traditional semantic communication faces limitations, including static coding strategies, poor generalization, and reliance on task-specific knowledge bases that hinder adaptability. To overcome these challenges, we propose a novel system combining scene understanding, Large Language Models (LLMs), and open channel coding, named \textbf{OpenSC}. Traditional systems rely on fixed domain-specific knowledge bases, limiting their ability to generalize. Our open channel coding approach leverages shared, publicly available knowledge, enabling flexible, adaptive encoding. This dynamic system reduces reliance on static task-specific data, enhancing adaptability across diverse tasks and environments. Additionally, we use scene graphs for structured semantic encoding, capturing object relationships and context to improve tasks like Visual Question Answering (VQA). Our approach selectively encodes key semantic elements, minimizing redundancy and improving transmission efficiency. Experimental results show significant improvements in both semantic understanding and efficiency, advancing the potential of adaptive, generalizable semantic communication in 6G networks.

Figures

Figures reproduced from arXiv: 2501.14520 by the authors.

Figure 1
Figure 1. The Comparison of a semantic communica￾tion framework with visual scene understanding, in￾corporating semantic coding and knowledge-agnostic channel coding, with traditional semantic communica￾tion approaches [2]. the scene’s semantics is crucial. 2) Static Knowledge Bases: Most systems rely on static, task-specific knowledge bases that hinder adaptability and generalization across diverse tasks and dynamic environm… view at source ↗
Figure 2
Figure 2. The overall framework of scene understanding-enabled semantic communication with open-channel [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Recall performance evaluation under different SNR levels in an AWGN channel using a 16QAM modulation [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: F1-score evaluation for four types of questions (Category, Quantity, Location, Relationship) under [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Recall performance evaluation for four types of questions (Category, Quantity, Location, Relationship) [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: F1-score evaluation for four types of questions (Category, Quantity, Location, Relationship) under [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Recall performance for four types of questions (Category, Quantity, Location, Relationship) under [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: F1 score for four types of questions (Category, Quantity, Location, Relationship) under different SNR [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Visualization comparing the differences between our method and other methods. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Large-Scale Model Enabled Semantic Communication Based on Robust Knowledge Distillation

    cs.LG 2025-08 conditional novelty 6.0 of 10

    A framework that combines neural architecture search and knowledge distillation to compress a ViT-B/16 teacher into a compact, channel-robust semantic encoder for image classification.

Reference graph

Works this paper leans on

49 extracted references · 37 canonical work pages · cited by 1 Pith paper

  1. [1]

    A mathematical theory of communication

    Claude Elwood Shannon. A mathematical theory of communication. The Bell system technical journal , 27(3):379–423, 1948

    1948

  2. [2]

    Semantic communications for future internet: Fundamentals, applications, and challenges

    Wanting Yang, Hongyang Du, Zi Qin Liew, Wei Yang Bryan Lim, Zehui Xiong, Dusit Niyato, Xuefen Chi, Xuemin Shen, and Chunyan Miao. Semantic communications for future internet: Fundamentals, applications, and challenges. IEEE Communications Surveys & Tutorials, 25(1):213–250, 2022

    work page 2022

  3. [3]

    Toward wisdom- evolutionary and primitive-concise 6g: A new paradigm of semantic communication networks

    Ping Zhang, Wenjun Xu, Hui Gao, Kai Niu, Xiaodong Xu, Xiaoqi Qin, Caixia Yuan, Zhijin Qin, Haitao Zhao, Jibo Wei, et al. Toward wisdom- evolutionary and primitive-concise 6g: A new paradigm of semantic communication networks. Engineering, 8:60–73, 2022

    work page 2022

  4. [4]

    From semantic communication to semantic-aware networking: Model, architecture, and open problems

    Guangming Shi, Yong Xiao, Yingyu Li, and Xuemei Xie. From semantic communication to semantic-aware networking: Model, architecture, and open problems. IEEE Communications Magazine, 59(8):44–50, 2021

    work page 2021

  5. [5]

    Deep learning enabled semantic communication systems

    Huiqiang Xie, Zhijin Qin, Geoffrey Ye Li, and Biing-Hwang Juang. Deep learning enabled semantic communication systems. IEEE Transactions on Signal Processing, 69:2663–2675, 2021

    work page 2021

  6. [6]

    Semantic communica- tions for speech signals

    Zhenzi Weng, Zhijin Qin, and Geoffrey Ye Li. Semantic communica- tions for speech signals. In ICC 2021 - IEEE International Conference on Communications, pages 1–6, 2021

    work page 2021

  7. [7]

    Deep learning-based image semantic coding for semantic communications

    Danlan Huang, Xiaoming Tao, Feifei Gao, and Jianhua Lu. Deep learning-based image semantic coding for semantic communications. In 2021 IEEE Global Communications Conference (GLOBECOM) , pages 1–6, 2021

    2021

  8. [8]

    Wireless deep video semantic trans- mission

    Sixian Wang, Jincheng Dai, Zijian Liang, Kai Niu, Zhongwei Si, Chao Dong, Xiaoqi Qin, and Ping Zhang. Wireless deep video semantic trans- mission. IEEE Journal on Selected Areas in Communications , 41(1):214– 229, 2023

    work page 2023

Show all 49 references
  1. [9]

    BERT: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of deep bidirectional transformers for language understanding. In Jill Burstein, Christy Doran, and Thamar Solorio, editors, Proceedings of the 2019 Conference of the North American Chap- ter ...

    2019

  2. [10]

    Improving language under- standing by generative pre-training

    Alec Radford and Karthik Narasimhan. Improving language under- standing by generative pre-training. 2018

    2018

  3. [11]

    Unbiased scene graph generation from biased training

    Kaihua Tang, Yulei Niu, Jianqiang Huang, Jiaxin Shi, and Hanwang Zhang. Unbiased scene graph generation from biased training. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3716–3725, 2020

    2020

  4. [12]

    Llm-enhanced composed image retrieval: An intent uncertainty- aware linguistic-visual dual channel matching model

    Hongfei Ge, Yuanchun Jiang, Jianshan Sun, Kun Yuan, and Yezheng Liu. Llm-enhanced composed image retrieval: An intent uncertainty- aware linguistic-visual dual channel matching model. ACM Trans. Inf. Syst., October 2024. Just Accepted

    2024

  5. [13]

    Mohan Raparthi, Venkata Siva Prakash Nimmagadda, Mohit Kumar Sahu, Swaroop Reddy Gayam, Sandeep Pushyamitra Pattyam, Kr- ishna Kanth Kondapaka, Bhavani Prasad Kasaraneni, Praveen Thuniki, Siva Sarana Kuna, and Sudharshan Putha. Real-time ai decision mak- ing in iot with quantu...

    2021

  6. [14]

    A survey on neural network interpretability

    Yu Zhang, Peter Tiňo, Aleš Leonardis, and Ke Tang. A survey on neural network interpretability. IEEE Transactions on Emerging Topics in Computational Intelligence, 5(5):726–742, 2021

    2021

  7. [15]

    Semantic communi- cations: Overview, open issues, and future research directions

    Xuewen Luo, Hsiao-Hwa Chen, and Qing Guo. Semantic communi- cations: Overview, open issues, and future research directions. IEEE Wireless Communications, 29(1):210–219, 2022

    2022

  8. [16]

    Toward semantic communications: Deep learning-based image semantic coding

    Danlan Huang, Feifei Gao, Xiaoming Tao, Qiyuan Du, and Jianhua Lu. Toward semantic communications: Deep learning-based image semantic coding. IEEE Journal on Selected Areas in Communications , 41(1):55–71, 2023

    2023

  9. [17]

    Semantic communication system based on semantic slice models propagation

    Chen Dong, Haotai Liang, Xiaodong Xu, Shujun Han, Bizhu Wang, and Ping Zhang. Semantic communication system based on semantic slice models propagation. IEEE Journal on Selected Areas in Communications, 41(1):202–213, 2022

    2022

  10. [18]

    Task-oriented multi-user semantic communications for vqa

    Huiqiang Xie, Zhijin Qin, and Geoffrey Ye Li. Task-oriented multi-user semantic communications for vqa. IEEE Wireless Communications Letters, 11(3):553–557, 2022

    2022

  11. [19]

    Task- oriented multi-user semantic communications.IEEE Journal on Selected Areas in Communications, 40(9):2584–2597, 2022

    Huiqiang Xie, Zhijin Qin, Xiaoming Tao, and Khaled B Letaief. Task- oriented multi-user semantic communications.IEEE Journal on Selected Areas in Communications, 40(9):2584–2597, 2022

    2022

  12. [20]

    Cross-modal semantic communications

    Ang Li, Xin Wei, Dan Wu, and Liang Zhou. Cross-modal semantic communications. IEEE Wireless Communications, 29(6):144–151, 2022

    2022

  13. [21]

    H. Nyquist. Certain factors affecting telegraph speed. The Bell System Technical Journal, 3(2):324–346, 1924

    1924

  14. [22]

    Scene graph refinement network for visual question answering

    Tianwen Qian, Jingjing Chen, Shaoxiang Chen, Bo Wu, and Yu-Gang Jiang. Scene graph refinement network for visual question answering. IEEE Transactions on Multimedia , 25:3950–3961, 2022

    2022

  15. [23]

    Visual question answering based on multimodal triplet knowledge accumuation

    Fengjuan Wang and Gaoyun An. Visual question answering based on multimodal triplet knowledge accumuation. In 2022 16th IEEE International Conference on Signal Processing (ICSP) , volume 1, pages 81–84, 2022

    2022

  16. [24]

    Knowledge blended open domain visual question answering using transformer

    Dipali Koshti, Ashutosh Gupta, and Mukesh Kalla. Knowledge blended open domain visual question answering using transformer. In 2023 Third International Conference on Artificial Intelligence and Smart En- ergy (ICAIS), pages 823–828, 2023

    2023

  17. [25]

    Transformer-based sparse encoder and answer decoder for visual question answering

    Longkun Peng, Gaoyun An, and Qiuqi Ruan. Transformer-based sparse encoder and answer decoder for visual question answering. In 2022 16th IEEE International Conference on Signal Processing (ICSP), volume 1, pages 120–123, 2022

    2022

  18. [26]

    Nine challenges in artificial intelligence and wireless communications for 6g

    Wen Tong and Geoffrey Ye Li. Nine challenges in artificial intelligence and wireless communications for 6g. IEEE Wireless Communications, 29(4):140–145, 2022

    2022

  19. [27]

    Towards a theory of se- mantic communication

    Jie Bao, Prithwish Basu, Mike Dean, Craig Partridge, Ananthram Swami, Will Leland, and James A Hendler. Towards a theory of se- mantic communication. In 2011 IEEE Network Science Workshop, pages 110–117. IEEE, 2011

    2011

  20. [28]

    6g networks: Be- yond shannon towards semantic and goal-oriented communications

    Emilio Calvanese Strinati and Sergio Barbarossa. 6g networks: Be- yond shannon towards semantic and goal-oriented communications. Computer Networks, 190:107930, 2021

    2021

  21. [29]

    Joint coding- modulation for digital semantic communications via variational au- toencoder

    Yufei Bo, Yiheng Duan, Shuo Shao, and Meixia Tao. Joint coding- modulation for digital semantic communications via variational au- toencoder. IEEE Transactions on Communications , 2024

    2024

  22. [30]

    Multimodal trustworthy semantic communication for audio-visual event localization

    Yuandi Li, Zhe Xiang, Fei Yu, Zhangshuang Guan, Hui Ji, Zhiguo Wan, and Cheng Feng. Multimodal trustworthy semantic communication for audio-visual event localization. arXiv preprint arXiv:2411.01991 , 2024

    2024 arXiv

  23. [31]

    Multimodal and multiuser semantic communications for channel-level information fusion

    Xuewen Luo, Ruobin Gao, Hsiao-Hwa Chen, Shuyi Chen, Qing Guo, and Ponnuthurai Nagaratnam Suganthan. Multimodal and multiuser semantic communications for channel-level information fusion. IEEE Wireless Communications, 31(2):117–125, 2024

    2024

  24. [32]

    Large ai model empowered multimodal semantic Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Zhe Xiang, Fei Yu, Quan Deng, Yuandi Li, Zhiguo Wan communications

    Feibo Jiang, Li Dong, Yubo Peng, Kezhi Wang, Kun Yang, Cunhua Pan, and Xiaohu You. Large ai model empowered multimodal semantic Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Zhe Xiang, Fei Yu, Quan Deng, Yuandi Li, Zhiguo Wan communications. IEEE Communications Magaz...

    2018

  25. [33]

    Large ai model-based semantic communications

    Feibo Jiang, Yubo Peng, Li Dong, Kezhi Wang, Kun Yang, Cunhua Pan, and Xiaohu You. Large ai model-based semantic communications. IEEE Wireless Communications, 31(3):68–75, 2024

    2024

  26. [34]

    Large language models empowered autonomous edge ai for connected intelligence

    Yifei Shen, Jiawei Shao, Xinjie Zhang, Zehong Lin, Hao Pan, Dong- sheng Li, Jun Zhang, and Khaled B Letaief. Large language models empowered autonomous edge ai for connected intelligence. IEEE Communications Magazine, 2024

    2024

  27. [35]

    Vqa: Visual ques- tion answering

    Stanislaw Antol, Aishwarya Agrawal, Jiasen Lu, Margaret Mitchell, Dhruv Batra, C Lawrence Zitnick, and Devi Parikh. Vqa: Visual ques- tion answering. In Proceedings of the IEEE international conference on computer vision, pages 2425–2433, 2015

    2015

  28. [36]

    Stacked attention networks for image question answering

    Zichao Yang, Xiaodong He, Jianfeng Gao, Li Deng, and Alex Smola. Stacked attention networks for image question answering. In Proceed- ings of the IEEE conference on computer vision and pattern recognition , pages 21–29, 2016

    2016

  29. [37]

    Prototype-based embedding network for scene graph generation

    Chaofan Zheng, Xinyu Lyu, Lianli Gao, Bo Dai, and Jingkuan Song. Prototype-based embedding network for scene graph generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 22783–22792, 2023

    2023

  30. [38]

    Aug: A new dataset and an efficient model for aerial image urban scene graph generation

    Yansheng Li, Kun Li, Yongjun Zhang, Linlin Wang, and Dingwen Zhang. Aug: A new dataset and an efficient model for aerial image urban scene graph generation. ArXiv, abs/2404.07788, 2024

    2024 arXiv

  31. [39]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. Mask r-cnn. In Proceedings of the IEEE international conference on computer vision, pages 2961–2969, 2017

    2017

  32. [40]

    Feature pyramid networks for object detection

    Tsung-Yi Lin, Piotr Dollár, Ross Girshick, Kaiming He, Bharath Har- iharan, and Serge Belongie. Feature pyramid networks for object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2117–2125, 2017

    2017

  33. [41]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 770–778, 2016

    2016

  34. [42]

    Faster r- cnn: Towards real-time object detection with region proposal net- works

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r- cnn: Towards real-time object detection with region proposal net- works. IEEE transactions on pattern analysis and machine intelligence , 39(6):1137–1149, 2016

    2016

  35. [43]

    Karagiannidis

    Shunpu Tang, Qianqian Yang, Lisheng Fan, Xianfu Lei, Arumugam Nallanathan, and George K. Karagiannidis. Contrastive learning-based semantic communications. IEEE Transactions on Communications , 72(10):6328–6343, 2024

    2024

  36. [44]

    Large language model enabled semantic communi- cation systems

    Zhenyi Wang, Li Zou, Shengyun Wei, Feifan Liao, Jia Zhuo, Haibo Mi, and Rongxuan Lai. Large language model enabled semantic communi- cation systems. arXiv preprint arXiv:2407.14112, 2024

    2024 arXiv

  37. [45]

    The llama 3 herd of models

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Ka- dian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024

    2024 arXiv

  38. [46]

    The claude 3 model family: Opus, sonnet, haiku

  39. [47]

    Gpt-4 technical report

    OpenAI Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, and et al. Gpt-4 technical report. 2023

    2023

  40. [48]

    Qwen technical report

    Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, and et al. Qwen technical report. ArXiv, abs/2309.16609, 2023

    2023 arXiv

  41. [49]

    Retrieval- augmented generation for large language models: A survey

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. Retrieval- augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997, 2023

    2023 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.