Pith. sign in

REVIEW 3 major objections 4 minor 37 references

FedNano: Toward Lightweight Federated Tuning for Pretrained Multimodal Large Language Models

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read FedNano claims federated tuning of multimodal LLMs works with the frozen LLM on the server and only small client-side adapters, cutting client storage by 95% and communication to 0.01% of parameters while matching or beating federated…

desk verdict The central communication-efficiency claim is undercut by the architecture's per-step embedding/gradient traffic, but the server-hosted-LLM idea is a real direction worth a serious look. read the letter →

arxiv 2506.14824 v1 pith:4OSD7ZWY submitted 2025-06-12 cs.LG cs.AIcs.MM

classification cs.LGcs.AIcs.MM
keywords federatedlearningmultimodallargelanguagemodelsparameter-efficientfine-tuninglow-rankadaptationserver-sidedeploymentnon-IIDdatavisualquestionansweringFishermerging
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

FedNano tries to make federated learning practical for multimodal large language models, which are too large to run on client devices. It claims that the LLM can be frozen and kept on the server while each client trains only a small module, NanoEdge, made of frozen modality encoders and trainable low-rank adapters. The paper reports that this cuts client-side storage by more than 95% and the number of transmitted parameters to 0.01% of the model, and that the approach matches or beats existing PEFT-based federated baselines on ScienceQA and IconQA. A reader should care because, if correct, it removes the main obstacle, client-side model size, to federated tuning of billion-parameter multimodal models.

What carries the argument

The load-bearing object is NanoEdge: a client-side module composed of frozen modality-specific encoders (a vision transformer and a text embedding layer), a frozen modality connector, and trainable NanoAdapters, one for vision and one for text, implemented as LoRA-style low-rank down- and up-projections external to the LLM. These adapters sit at the boundary between the connector and the server-hosted LLM, so clients never execute the LLM. Aggregation is carried by Fisher merging, which weights each client's adapter updates by a diagonal Fisher Information Matrix approximated from squared gradients, serving as the precision matrix in a Laplace approximation of the client posterior and reducing the cost from $O(|\theta|^2)$ to $O(|\theta|)$.

What would settle it

Run FedNano and measure total bytes exchanged per round, including the embedding tensors clients must send for the server's frozen LLM forward pass and the gradient tensors returned, then compare that per-round byte count with FedDPA-F's transmitted adapter updates. If the embedding and gradient transfer exceeds the compared method's uploads, the 0.01% communication claim does not hold.

Watch

Extended reading notes

Core claim

The authors' claim is that adaptation can be decoupled from the LLM itself: instead of inserting adapters into the model and running it on each client, FedNano places the frozen LLM on the server and attaches lightweight NanoAdapters at the connector interface on the client. Only these adapter updates are aggregated, using Fisher-weighted averaging to handle non-IID data. On the paper's own experiments, FedNano reaches 77.05% average on ScienceQA and 76.42% on IconQA with MiniGPT-4, and 81.41% and 78.04% with LLaVA-1.5, exceeding FedAvg, FedProx, and FedDPA-F in each case, while using far fewer client resources.

Load-bearing premise

The central efficiency claim rests on counting only the tiny adapter updates as communication, while ignoring the per-step embeddings and gradients that must travel between client and server at every training step, and that unmeasured traffic could dwarf the advertised 0.01%.

Editorial extensions

If this is right

  • Federated tuning of MLLMs no longer requires shipping or executing the LLM on clients, so devices with limited storage and compute can participate.
  • Per-round communication drops to the size of the adapter updates (0.01% of model parameters), making low-bandwidth federated settings more plausible.
  • Under non-IID splits, including strongly heterogeneous Dirichlet partitions and a cross-task four-client setup, the Fisher-guided aggregation reports better average accuracy than FedAvg, FedProx, and FedDPA-F.
  • The dual-modality design matters: using either the visual or textual adapter alone yields clearly worse results than using both.
  • Accuracy improves with higher adapter rank and more frequent communication rounds, at the cost of more transmitted data.

Reading between the lines

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

  • The paper does not measure the per-step embedding and gradient traffic that the server-hosted LLM requires; adding that traffic would be needed to validate the 0.01% communication claim.
  • Because clients send adapted embeddings to the server, the privacy guarantee is structural rather than cryptographic, and stronger protection would likely require differential privacy or encryption.
  • The adapter interface is modular enough that the same scheme could extend to other modality pairs or to server-hosted models that expose only an embedding API, since the adapters never touch internal LLM weights.
  • The FedNano-EF variant indicates most of the accuracy gains survive without the extra FIM computation, suggesting a cost-accuracy trade-off that the main paper does not emphasize.
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. The paper proposes FedNano, a federated tuning framework for multimodal LLMs in which the frozen LLM is hosted on the server and clients train only a lightweight 'NanoEdge' module consisting of frozen modality encoders, a connector, and low-rank adapters (AI for images, AT for text). Aggregation uses a Fisher-information-weighted average (Eq. 1). The authors claim a 95.7% reduction in client-side storage, a 99.4% reduction in transmitted parameters relative to FedDPA-F, and accuracy gains over FedAvg/FedProx/FedDPA-F on ScienceQA and IconQA with MiniGPT-4 and LLaVA-1.5. The paper also reports ablations on data heterogeneity level, client count, cross-task distributions, adapter rank, and communication frequency.

Significance. The architectural idea of centralizing the large backbone on the server and using external, client-side adapters so that clients never store or execute the LLM is a reasonable and potentially useful direction for MLLM federated learning, provided the communication cost is honestly measured. The empirical evaluation uses external benchmarks (ScienceQA, IconQA) and standard FL baselines, so the claimed gains are not circular. However, the headline efficiency result is based on an incomplete measure of communication, and the accuracy comparisons are too weak, with small margins, no error bars, and very small scale, to support the paper's conclusions. With those caveats, the contribution is not yet established; the paper also provides no code or detailed system measurements to support the resource-efficiency claims.

major comments (3)
  1. [Abstract; Section 3.2; Table 1; Figure 2] The headline communication-efficiency claim is based on an incomplete accounting. The 0.01% figure and the statement that 'only compact NanoAdapter updates are exchanged' count solely the NanoAdapter parameter deltas (1.05M parameters per round in Table 1) and omit the per-local-step transfers that the architecture in Figure 2 and Algorithm 1 requires: each client must upload adapted visual and text embeddings to the server for the frozen LLM forward pass and receive gradients back for backpropagation through the adapters. For LLaVA-1.5-7B, the projected visual embedding alone is 576 tokens times 4096 dimensions, which is about 9.4 MB per sample in fp32, so with batch size 8 the per-step bidirectional traffic is already on the order of 150 MB; over 10 rounds with hundreds to thousands of local samples this can dominate or exceed the roughly 723 MB per round that FedDPA-F transmits as parameter updates. The paper's own Limitation section repeats the claim that FedNano transmits 'only lightweight adapter updates,' which is contradicted by the architecture. Since the central contribution is suitability for bandwidth-constrained environments, this accounting error invalidates the main efficiency result as stated.
  2. [Section 4.2; Tables 2 and 3] The accuracy comparisons do not support the claim that FedNano 'outperforms' baselines. All numbers are single-seed runs with 5 clients, 10 rounds, and 1 local epoch per round, and the margins are small: on LLaVA-1.5, FedNano exceeds FedAvg by 0.65 points on ScienceQA and 0.32 points on IconQA (Table 2), and at alpha = 5 it leads FedAvg by 0.03 points (Table 3). Without error bars, repeated seeds, or a significance test, these differences are within run-to-run noise for this setting, so the cross-method ranking in Tables 2-5 is not established.
  3. [Section 3.4, Eq. (1)] Equation (1) as written does not do dataset-size weighting: the factor |D_k|/sum|D_k| appears in both the numerator and the denominator and cancels, leaving theta_global = (sum F_k theta_k)/(sum F_k). The surrounding text describes the method as combining posterior estimates with dataset-size weighting; the formula should be corrected, and the claimed benefit of client-data-volume weighting should be verified or removed. This does not by itself invalidate the aggregation idea, but it makes the description in Section 3.4 inaccurate.
minor comments (4)
  1. [Algorithm 1] The FIM computation is specified only as 'Compute FIM F_k'; Section 3.4 mentions a diagonal approximation from squared gradients, but the exact computation, including which parameters are included and how many samples are used, should be stated for reproducibility.
  2. [Figure 2] Figure 2 shows 'Client to Server' and 'Server to Client' arrows that appear to represent per-step activation and gradient transfers, but the caption only describes adapter uploads; the communication protocol should be described precisely in the caption or text.
  3. [Section 4.2] The paper does not state the number of random seeds or show variance for any experiment; adding this information is necessary to support the reported rankings.
  4. [Throughout] Several typos and spacing errors remain, such as 'F edNano' and 'LLaV A-1.5' in the text; these should be cleaned up in revision.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the paper's accuracy claims are externally benchmarked, and its efficiency numbers are parameter-count arithmetic rather than predictions derived from fitted inputs.

full rationale

FedNano's load-bearing results are empirical comparisons on ScienceQA and IconQA against FedAvg, FedProx, and FedDPA-F, using external benchmarks and standard baselines. No fitted parameter is relabeled as a prediction: the Fisher-merging aggregation (Eq. 1) is attributed to external prior work (Matena and Raffel 2022), and the method's effectiveness is judged by held-out accuracy. The advertised 0.01% communication and 95% storage figures in Table 1 are arithmetic consequences of the chosen parameter counts, not derived predictions; the disputed omission of per-step embedding/gradient transfers between client and server is a metric-accounting or correctness concern, not a circular dependency in the derivation. The self-citations to Chen et al. 2023 (FedDat) appear in related-work and assumption statements such as 'we assume complete modality availability and a shared model architecture across all clients; only data distributions differ (Chen et al., 2023)' and are not load-bearing: the PEFT-FL characterization is also attributed to external works (Wang et al. 2024; Zhang et al. 2024), and the assumption is a standard FL setup condition rather than a result that FedNano derives from itself. The Section 4.2 novelty assertion ('FedNano is the first FL framework...') is a claim of absence of prior work, not a mathematical derivation, and does not constitute circularity. No uniqueness theorem, ansatz, or fitted quantity is smuggled in through a self-citation chain. Overall score reflects one minor, non-load-bearing self-citation; there is no substantive circularity.

Assumptions & free parameters 1 free parameters · 4 assumptions · 2 invented entities

The paper introduces two named components (NanoEdge, NanoAdapters) but they are compositions of known building blocks rather than conceptually new entities with independent evidence. The main assumptions are about modality availability, server trust, diagonal FIM adequacy, and the benefit of freezing encoders and connector. The only hand-chosen numeric quantity that directly shapes the communication claims is the adapter rank.

free parameters (1)
  • NanoAdapter rank = 64 (default; 16,32,128,256 in ablation)
    The rank of the low-rank adapter determines capacity and communication size; the paper selects rank 64 for the main tables and shows accuracy varies with rank in Fig. 3b.
assumptions (4)
  • domain assumption All clients have complete modality availability and a shared model architecture.
    Stated in Section 3.1; the limitation about incomplete modalities is acknowledged in Section 5.
  • domain assumption The server can process client embeddings and gradients during training without breaking privacy.
    FedNano's training requires the server to run the frozen LLM on client embeddings; the paper assumes this is privacy-preserving without a threat model.
  • domain assumption The diagonal Fisher information matrix computed from squared gradients is a sufficient importance weight for aggregation.
    Used in Eq. (1) following Matena and Raffel; the paper adopts the Laplace approximation and diagonal approximation without validation on these models.
  • domain assumption Freezing the modality encoders and connector preserves the pretrained alignment with the frozen LLM.
    Core design hypothesis used in Section 3.3; the paper does not test alternative placements or unfreezing.
invented entities (2)
  • NanoEdge independent evidence
    purpose: Client-side module containing frozen encoders, connector, and trainable NanoAdapters.
    Empirical results in Tables 2-6 support its usefulness, but it is composed entirely of existing components (LoRA, frozen encoders).
  • NanoAdapter (vision and text) independent evidence
    purpose: Low-rank adapters placed at the connector-LLM interface to adapt representations without structural access to the LLM.
    Ablation in Table 6 shows both adapters contribute, but the mechanism is the same as LoRA applied externally.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FedNano: Toward Lightweight Federated Tuning for Pretrained Multimodal Large Language Models." pith.science (2026). https://pith.science/paper/4OSD7ZWY

@misc{pith2026250614824,
  author       = {Pith},
  title        = {Pith review of: FedNano: Toward Lightweight Federated Tuning for Pretrained Multimodal Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4OSD7ZWY}},
  note         = {Machine review of arXiv:2506.14824}
}
read the original abstract

Multimodal Large Language Models (MLLMs) excel in tasks like multimodal reasoning and cross-modal retrieval but face deployment challenges in real-world scenarios due to distributed multimodal data and strict privacy requirements. Federated Learning (FL) offers a solution by enabling collaborative model training without centralizing data. However, realizing FL for MLLMs presents significant challenges, including high computational demands, limited client capacity, substantial communication costs, and heterogeneous client data. Existing FL methods assume client-side deployment of full models, an assumption that breaks down for large-scale MLLMs due to their massive size and communication demands. To address these limitations, we propose FedNano, the first FL framework that centralizes the LLM on the server while introducing NanoEdge, a lightweight module for client-specific adaptation. NanoEdge employs modality-specific encoders, connectors, and trainable NanoAdapters with low-rank adaptation. This design eliminates the need to deploy LLM on clients, reducing client-side storage by 95%, and limiting communication overhead to only 0.01% of the model parameters. By transmitting only compact NanoAdapter updates, FedNano handles heterogeneous client data and resource constraints while preserving privacy. Experiments demonstrate that FedNano outperforms prior FL baselines, bridging the gap between MLLM scale and FL feasibility, and enabling scalable, decentralized multimodal AI systems.

Figures

Figures reproduced from arXiv: 2506.14824 by the authors.

Figure 1
Figure 1. Comparison between traditional PEFT-based [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the FedNano framework. The server hosts the frozen LLM, while each client performs local tuning via NanoEdge, which includes NanoAdapter-T for text and NanoAdapter-I for vision. Clients upload low-rank adapter updates, which are aggregated on the server using Fisher merging. This design reduces client overhead and supports scalable, multimodal federated learning under data heterogeneity. tion and enablin… view at source ↗
Figure 3
Figure 3. (a) Impact of communication frequency. Fed￾Nano outperforms FedAvg, with more frequent commu￾nication amplifying its advantages; (b) Effect of adapter rank. FedNano consistently achieves superior perfor￾mance, demonstrating its ability to capture task-specific and client-specific information effectively. nificantly enhances the effectiveness of FedNano, reinforcing its superior ability to integrate client￾specific u… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 8 canonical work pages

  1. [1]

    Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katherine Millican, Malcolm Reynolds, et al. 2022. Flamingo: a visual language model for few-shot learning. Advances in neural information processing systems, 35:23716--23736

  2. [2]

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

  3. [3]

    Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, et al. 2023. Qwen technical report. arXiv preprint arXiv:2309.16609

  4. [4]

    Liwei Che, Jiaqi Wang, Xinyue Liu, and Fenglong Ma. 2024. https://arxiv.org/abs/2406.11048 Leveraging foundation models for multi-modal federated learning with incomplete modality . Preprint, arXiv:2406.11048

  5. [5]

    Tianshi Che, Ji Liu, Yang Zhou, Jiaxiang Ren, Jiwen Zhou, Victor S Sheng, Huaiyu Dai, and Dejing Dou. 2023. Federated learning of large language models with parameter-efficient prompt tuning and adaptive optimization. arXiv preprint arXiv:2310.15080

  6. [6]

    Haokun Chen, Yao Zhang, Denis Krompass, Jindong Gu, and Volker Tresp. 2023. https://arxiv.org/abs/2308.12305 Feddat: An approach for foundation model finetuning in multi-modal heterogeneous federated learning . Preprint, arXiv:2308.12305

  7. [7]

    Jiayi Chen and Aidong Zhang. 2024. On disentanglement of asymmetrical knowledge transfer for modality-task agnostic federated learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 11311--11319

  8. [8]

    Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale Fung, and Steven Hoi. 2023. https://arxiv.org/abs/2305.06500 Instructblip: Towards general-purpose vision-language models with instruction tuning . Preprint, arXiv:2305.06500

Show all 37 references
  1. [9]

    Tiantian Feng, Digbalay Bose, Tuo Zhang, Rajat Hebbar, Anil Ramakrishna, Rahul Gupta, Mi Zhang, Salman Avestimehr, and Shrikanth Narayanan. 2023. https://arxiv.org/abs/2306.09486 Fedmultimodal: A benchmark for multimodal federated learning . Preprint, arXiv:2306.09486

  2. [10]

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. 2019. Parameter-efficient transfer learning for nlp. In International conference on machine learning, pages 2790--2799. PMLR

  3. [11]

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685

  4. [12]

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. 2017. Overcoming catastrophic forgetting in neural networks. Proceedings of the national academy of...

  5. [13]

    Fan Lai, Yinwei Dai, Sanjay Singapuram, Jiachen Liu, Xiangfeng Zhu, Harsha Madhyastha, and Mosharaf Chowdhury. 2022. Fedscale: Benchmarking model and system performance of federated learning at scale. In International conference on machine learning, pages 11814--11827. PMLR

  6. [14]

    Brian Lester, Rami Al-Rfou, and Noah Constant. 2021. The power of scale for parameter-efficient prompt tuning. arXiv preprint arXiv:2104.08691

  7. [15]

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. 2023. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning, pages 19730--19742. PMLR

  8. [16]

    Tian Li, Anit Kumar Sahu, Manzil Zaheer, Maziar Sanjabi, Ameet Talwalkar, and Virginia Smith. 2020. Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems, 2:429--450

  9. [17]

    Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. 2024 a . Improved baselines with visual instruction tuning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 26296--26306

  10. [18]

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2024 b . Visual instruction tuning. Advances in neural information processing systems, 36

  11. [19]

    Pan Lu, Swaroop Mishra, Tony Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Oyvind Tafjord, Peter Clark, and Ashwin Kalyan. 2022. Learn to explain: Multimodal reasoning via thought chains for science question answering. In The 36th Conference on Neural Information Processing Sy...

  12. [20]

    Pan Lu, Liang Qiu, Jiaqi Chen, Tony Xia, Yizhou Zhao, Wei Zhang, Zhou Yu, Xiaodan Liang, and Song-Chun Zhu. 2021. Iconqa: A new benchmark for abstract diagram understanding and visual language reasoning. In The 35th Conference on Neural Information Processing Systems (NeurIPS)...

  13. [21]

    Michael S Matena and Colin A Raffel. 2022. Merging models with fisher-weighted averaging. Advances in Neural Information Processing Systems, 35:17703--17716

  14. [22]

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. 2017. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, pages 1273--1282. PMLR

  15. [23]

    Baolin Peng, Chunyuan Li, Pengcheng He, Michel Galley, and Jianfeng Gao. 2023 a . Instruction tuning with gpt-4. arXiv preprint arXiv:2304.03277

  16. [24]

    Zhiliang Peng, Wenhui Wang, Li Dong, Yaru Hao, Shaohan Huang, Shuming Ma, and Furu Wei. 2023 b . Kosmos-2: Grounding multimodal large language models to the world. arXiv preprint arXiv:2306.14824

  17. [25]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth \'e e Lacroix, Baptiste Rozi \`e re, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971

  18. [26]

    Ziyao Wang, Zheyu Shen, Yexiao He, Guoheng Sun, Hongyi Wang, Lingjuan Lyu, and Ang Li. 2024. Flora: Federated fine-tuning large language models with heterogeneous low-rank adaptations. arXiv preprint arXiv:2409.05976

  19. [27]

    Chengyue Wu, Teng Wang, Yixiao Ge, Zeyu Lu, Ruisong Zhou, Ying Shan, and Ping Luo. 2023. -tuning: transferring multimodal foundation models with optimal multi-task interpolation. In Proceedings of the 40th International Conference on Machine Learning, pages 37713--37727

  20. [28]

    Binqian Xu, Xiangbo Shu, Haiyang Mei, Guosen Xie, Basura Fernando, Mike Zheng Shou, and Jinhui Tang. 2024. Fedmllm: Federated fine-tuning mllm on multimodal heterogeneity data. arXiv preprint arXiv:2411.14717

  21. [29]

    Yiyuan Yang, Guodong Long, Tao Shen, Jing Jiang, and Michael Blumenstein. 2024. Dual-personalizing adapter for federated foundation models. arXiv preprint arXiv:2403.19211

  22. [30]

    Liping Yi, Han Yu, Gang Wang, and Xiaoguang Liu. 2023. Fedlora: Model-heterogeneous personalized federated learning with lora tuning. arXiv preprint arXiv:2310.13283

  23. [31]

    Shukang Yin, Chaoyou Fu, Sirui Zhao, Ke Li, Xing Sun, Tong Xu, and Enhong Chen. 2024. A survey on multimodal large language models. National Science Review, page nwae403

  24. [32]

    Qiying Yu, Yang Liu, Yimu Wang, Ke Xu, and Jingjing Liu. 2023. https://arxiv.org/abs/2302.08888 Multimodal federated learning via contrastive representation ensemble . Preprint, arXiv:2302.08888

  25. [33]

    Elad Ben Zaken, Shauli Ravfogel, and Yoav Goldberg. 2021. Bitfit: Simple parameter-efficient fine-tuning for transformer-based masked language-models. arXiv preprint arXiv:2106.10199

  26. [34]

    Jianyi Zhang, Saeed Vahidian, Martin Kuo, Chunyuan Li, Ruiyi Zhang, Tong Yu, Guoyin Wang, and Yiran Chen. 2024. Towards building the federatedgpt: Federated instruction tuning. In ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP...

  27. [35]

    Deyao Zhu, Jun Chen, Xiaoqian Shen, Xiang Li, and Mohamed Elhoseiny. 2023. Minigpt-4: Enhancing vision-language understanding with advanced large language models. arXiv preprint arXiv:2304.10592

  28. [36]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  29. [37]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.