Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

NVCiM-PT: An NVCiM-assisted Prompt Tuning Framework for Edge LLMs

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Edge LLMs can personalize to each user by storing per-domain prompt vectors in non-volatile computing-in-memory arrays and retrieving the right prompt with an in-memory similarity search, a scheme the paper reports lifts accuracy by up to…

desk verdict A plausible but under-validated system paper: the NVCiM-assisted prompt tuning idea is real and the empirical sweep is broad, but SSA retrieval needs oracle/random controls before the headline claim is proven. read the letter →

arxiv 2411.08244 v1 pith:AKSA75TD submitted 2024-11-12 cs.LG cs.ET

Ruiyang Qin , Pengyu Ren , Zheyu Yan , Liu Liu , Dancheng Liu , Amir Nassereldine , Jinjun Xiong , Kai Ni
show 2 more authors
Sharon Hu Yiyu Shi
This is my paper · ORCID
classification cs.LGcs.ET
keywords prompttuningedgeLLMnon-volatilecomputing-in-memoryvirtualtokensdomainshiftdevicevariationpersonalizedin-memorysearch
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

This paper argues that the domain-shift problem that cripples prompt tuning on edge LLMs can be solved by storing many small, specialized prompt vectors—one per user-data cluster—inside non-volatile computing-in-memory (NVCiM) arrays and retrieving the right one for each new input with an in-memory similarity search. The proposed framework, NVCiM-PT, contributes three pieces: a k-means-based representative selection that decides which user samples deserve their own prompt, a noise-aware training scheme that makes prompts robust to the conductance variations of real memory devices, and a scaled-search algorithm that pools user inputs and stored prompts at multiple granularities and computes a weighted multi-scale dot product to pick the best prompt. The paper reports that this raises accuracy by up to 36.7 percent over existing noise-mitigation and retrieval baselines across five datasets, three edge-scale LLMs, and five memory-device models, while cutting retrieval latency by up to 120x and energy by up to 60x compared with a typical edge CPU. If true, this would make per-user personalization of on-device LLMs practical under severe memory and power budgets.

What carries the argument

The load-bearing object is the optimal set of virtual tokens (OVT), a small learned prompt vector specialized to a particular user-data domain rather than shared across all inputs. The mechanism that carries the argument is the store-and-scaled-search algorithm (SSA) on NVCiM: each OVT is encoded via a pretrained autoencoder, stored in a crossbar array, and retrieved by evaluating Eq. 5, the weighted multi-scale dot product (WMSDP), which average-pools both the user input embedding and the stored prompt at scales 1, 2, and 4, weights the three dot products, and selects the OVT with the highest score. The same crossbar array that stores the prompts also performs the matrix-matrix multiplications for the search, which is what turns a memory-size bottleneck into a latency and energy win. Around this sit the representative-selection step (k-means clustering plus cosine-similarity sample choice) and noise-aware training (Gaussian noise injection scaled by prompt magnitude) that make the OVTs worth storing in the first place.

What would settle it

Take one held-out user, train an OVT on that user's own data, and compare the LLM's accuracy with that oracle prompt against the accuracy with the prompt that SSA retrieves for the same user; if the SSA-retrieved prompt's accuracy is no better than a randomly chosen prompt's, the retrieval score is not measuring prompt quality.

Watch

Extended reading notes

Core claim

The paper's central claim is that the optimal set of virtual tokens (OVT) idea—previously restricted to cloud settings—can be brought to edge LLMs if the per-sample prompts are stored and retrieved inside NVCiM. Rather than training one 'one4all' prompt on all users, NVCiM-PT trains separate prompts on representative samples from each detected data domain, stores them encoded as low-precision vectors in non-volatile crossbar arrays, and at inference time retrieves the most appropriate prompt by running a weighted multi-scale dot product between pooled user-input embeddings and pooled stored prompts as in-situ matrix-matrix multiplications. Noise-aware training injects device-variation noise into the prompts during optimization so the stored prompts tolerate the conductance fluctuations of RRAM and FeFET devices. The upshot is that prompt tuning's accuracy advantage and its low trainable-parameter count are retained, while the storage and retrieval costs that once made per-user prompts impractical are absorbed by the memory hardware itself.

Load-bearing premise

The framework assumes that pooling the user input and a stored prompt at a few scales and taking their weighted dot product reliably identifies the prompt that will actually improve the LLM, even though word embeddings and virtual tokens live in different representational spaces; if pooling does not bridge that gap, the retrieved prompt is no better than a random one.

Editorial extensions

If this is right

  • Per-user personalization no longer requires retraining or large adapter storage: the edge LLM stays frozen and only the prompt changes.
  • The prompt library for a device is bounded by NVM capacity, not DRAM, because every stored prompt is searched by the same in-memory matrix multiplication.
  • The reported 36.7 percent accuracy gain indicates that a large share of prompt tuning's domain-shift loss is a retrieval problem rather than a training-capacity ceiling.
  • All prompts in the library are searched in parallel in the crossbar, so retrieval latency grows only weakly (if at all) as the number of stored user domains increases.

Reading between the lines

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

  • A natural extension: if the multi-scale pooling truly bridges word-embedding and virtual-token spaces, the same retrieval scheme could choose among other parameter-efficient fine-tuning artifacts such as LoRA adapters stored in memory.
  • The buffer-size sweep's intermediate optimum suggests an adaptive policy—grow the prompt library only while marginal retrieval quality keeps improving—but the paper does not derive such a policy.
  • The reported 120x and 60x gains cover the retrieval search itself; an end-to-end measurement that includes writing new prompts and updating the autoencoder would show how much of the gain survives a full personalization cycle.
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

5 major / 5 minor

Summary. The paper proposes NVCiM-PT, a prompt-tuning framework for edge LLMs that leverages non-volatile computing-in-memory (NVCiM) hardware. The framework has three components: representative selection (RS) to reduce the data buffer via k-means clustering and cosine-similarity selection; noise-aware training (NT) that injects Gaussian device-variation noise into virtual tokens during prompt tuning; and a store-and-scaled-search algorithm (SSA) that encodes virtual tokens with an autoencoder, stores them in NVM crossbars, and retrieves an OVT for each user input using a weighted multi-scale dot product. Experiments across three LLMs, five LaMP datasets, five NVM devices, and five baselines report consistent improvements, with up to 36.7% accuracy/ROUGE gains, 120x lower search latency, and 60x lower energy than a Jetson Orin CPU. The central claim is that NVCiM-PT, and in particular the SSA retrieval, enables effective use of optimal virtual tokens (OVTs) for domain-shift adaptation under edge resource constraints.

Significance. If the retrieval and noise-mitigation effects are real, NVCiM-PT addresses a genuine problem: improving edge LLM personalization under domain shift with limited resources by storing per-domain prompts in NVM. The empirical scope is broad: three LLMs, five task datasets, five NVM devices, and multiple baselines, with consistent gains in Tables I, III, and IV. The framework is a plausible co-design of prompt tuning and CiM architecture. However, the paper's main novelty, SSA, is only compared against MIPS and not against random or oracle retrieval, so the attribution of gains to SSA's ranking quality is not established. The missing hyperparameter values and lack of variance reporting further weaken the reproducibility and statistical confidence. The latency/energy results are promising but preliminary because they rely on NeuroSim simulations without a clear end-to-end accounting. With additional controls and specifications, the contribution could be solid.

major comments (5)
  1. [§III-D2, Eq. (5)] The SSA retrieval ranking is load-bearing for the paper's central claim, but it is never validated in isolation. The experiments compare SSA against MIPS and noise-mitigation baselines, yet they omit two essential controls: a random-OVT baseline (lower bound) and an oracle-retrieval baseline where the OVT is known to correspond to the test user/sample (upper bound). Without these, the gains in Tables I, III, and IV could be produced entirely by the representative selection and noise-aware training, while WMSDP might be retrieving near-random prompts. Since Section III-D2 itself acknowledges that word embeddings and virtual tokens are not compatible, the pooling assumption in Eq. (5) needs direct validation. Please add experiments showing that WMSDP ranks OVTs by their actual benefit to the LLM, e.g., by comparing the LLM performance under SSA-selected OVTs with random and oracle OVTs across several datasets and devices.
  2. [§III-B, Eq. (3)] Equation (3) uses argmin of cosine similarity to select the most representative sample e*_i in each cluster. Taken literally, argmin selects the sample least similar to the cluster centroid, which is the opposite of representativeness. This appears to be a typographical error for argmax, but the discrepancy must be resolved because the RS component is the first step of the pipeline and directly affects which OVTs are trained. If argmin was intended, the authors should justify why outlier samples are beneficial.
  3. [§III-B, Eq. (2) and §III-C, Eq. (4)] Key hyperparameters in the proposed method are not specified, making the results irreproducible. Equation (2) requires values for nmin, nmax, b0, and s, while Eq. (4) requires the noise factors f1, f2, f3, f4 and the relationship between the global σ (described as 0.1 in Section IV-A2) and the level-dependent standard deviations in Table II. None of these values are reported in the experimental setup or in the default settings. Please provide the exact values used for all reported experiments, or include a sensitivity study showing how performance varies with these parameters.
  4. [Tables III and IV] No repeated-run variance or confidence intervals are reported. The experiments use small buffer sizes (10–60 samples) and stochastic training, so differences such as 0.207 vs. 0.215 in Table III or 0.215 vs. 0.206 in Table IV may not be statistically significant. Please report the mean and standard deviation (or confidence intervals) across at least three independent runs, and indicate whether the observed improvements over the best baseline are statistically significant.
  5. [§IV-B2, Fig. 5] The latency and energy claims (up to 120x and 60x) are based on NeuroSim simulations for NVCiM crossbars compared with a Jetson Orin CPU. It is unclear whether the simulation and the CPU measurement account for the same end-to-end retrieval workload, including the overhead of encoding/decoding the query, data movement, and the LLM inference itself. Please clarify the scope of the measurement and report the speedup/energy ratio for the complete search pipeline rather than only the crossbar-array operation; otherwise the headline speedup numbers may overstate the benefit.
minor comments (5)
  1. [§III-D2, Eq. (5)] The equation for WMSDP is malformed: the denominator 'P i ∈Lwi' is written inline and the fraction is ambiguous. Please rewrite it as a properly typeset fraction and clarify the indexing over the scale set L.
  2. [§III-D2, pooling definition] The definition 'Pool_i(x) = 1/i \sum_{j=1}^{i} x_k : k + i' is hard to parse. Please provide a precise definition with explicit index ranges, e.g., average pooling over non-overlapping windows of size i.
  3. [§IV-A2 and Table II] In Section IV-A2 the text refers to 'TABE II' (a typo for Table II) and states that the device variation σ is 0.1, but Table II gives per-level variation values that differ across devices. Clarify how the global σ=0.1 relates to the per-device, per-level values used in the experiments.
  4. [Fig. 3] The framework overview in Fig. 3 is dense and the three components (RS, NT, SSA) are not clearly separated in the caption or the figure labels. Adding named bounding boxes or distinct colors with a legend would improve readability.
  5. [Abstract and Introduction] The phrase 'NVCiM' is introduced as 'non-volatile computing-in-memory' but later used as a noun ('NVCiM devices', 'NVCiM-PT'). Consider defining the abbreviation once at first use and using it consistently throughout.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the performance claims are empirical measurements, and no load-bearing argument reduces to its own inputs by construction.

full rationale

The paper's central claims are empirical: improved accuracy/ROUGE scores, latency, and energy are measured outcomes from experiments, not derived from the framework's own parameters. The OVT advantage is supported by an explicit independent comparison in Fig. 1, where prefix tuning with OVTs outperforms Vanilla, DEPT, and P-tuning v2 across models and datasets; the self-citation to [13] for the OVT concept is therefore not load-bearing because the paper re-derives the effect empirically. The SSA retrieval algorithm (Eq. 5) is a manually specified heuristic with fixed weights (W1,W2,W3 = 1, 0.8, 0.6); these weights are not fitted to the reported accuracy targets, so no fitted input is renamed as a prediction. Noise-aware training does inject a Gaussian noise distribution with the same sigma used in some evaluations, but this is standard robustness training, not a tautology: the paper also sweeps device variation sigma in Table IV, and the method does not bake the reported accuracy values into the training objective. The representative-selection equation (Eq. 3) contains an apparent argmin/argmax typo, but this is a correctness issue, not a circularity. No step in the derivation chain reduces to its own output by definition, and no argument is forced solely by a self-citation chain. Accordingly, the paper exhibits no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The framework relies on several unstated or under-specified design choices: the cluster count formula, noise factors, and SSA weights are effectively free parameters. The core assumptions about OVT superiority, Gaussian device noise, and the validity of pooling for retrieval are borrowed from prior work or introduced ad hoc. No new physical entities are proposed.

free parameters (6)
  • nmin, nmax, b0, s (k-means cluster count)
    Equation 2 determines the number of clusters k; no values are reported in the paper.
  • f1, f2, f3, f4 (noise factors)
    Equation 4 applies per-interval noise scaling; the factors are never specified.
  • sigma (noise standard deviation) = 0.1 (default), varied 0.025-0.15
    Global noise parameter for NVM variation; used as a fixed input in most experiments.
  • W1, W2, W3 (SSA weights) = 1, 0.8, 0.6
    Weights for scale-1,2,4 pooling in WMSDP; chosen by hand, no sensitivity study.
  • Scale factors L = {1,2,4} = 1, 2, 4
    Average pooling scales; fixed design choice.
  • Autoencoder size/precision = embedding 48, int16, subArray 384x128, 2-bit devices
    Storage configuration; no ablation on how these affect retrieval accuracy.
assumptions (5)
  • domain assumption OVTs trained per data sample or domain outperform a single one4all prompt set
    The motivation for the framework (Fig. 1); attributed to prior work [13] but treated as given.
  • domain assumption Device variation can be modeled as zero-mean Gaussian noise with per-level standard deviations
    Used in Eq. 4 and throughout; based on measurements in [20], [21].
  • domain assumption A pre-trained autoencoder can encode virtual tokens to int16 representations and decode them without meaningful loss for retrieval and generation
    The autoencoder is pre-trained but no architecture or training details are given (Sec. III-D1).
  • ad hoc to paper Average pooling at scales 1, 2, 4 makes the dot product between input embeddings and virtual tokens a meaningful similarity measure
    The incompatibility is acknowledged; the multi-scale pooling is introduced to fix it, but no theoretical or empirical justification is provided beyond the main results.
  • standard math k-means with cosine similarity accurately separates user data into domains
    Standard clustering applied to embeddings; assumes embeddings are meaningful.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NVCiM-PT: An NVCiM-assisted Prompt Tuning Framework for Edge LLMs." pith.science (2026). https://pith.science/paper/AKSA75TD

@misc{pith2026241108244,
  author       = {Pith},
  title        = {Pith review of: NVCiM-PT: An NVCiM-assisted Prompt Tuning Framework for Edge LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AKSA75TD}},
  note         = {Machine review of arXiv:2411.08244}
}
read the original abstract

Large Language Models (LLMs) deployed on edge devices, known as edge LLMs, need to continuously fine-tune their model parameters from user-generated data under limited resource constraints. However, most existing learning methods are not applicable for edge LLMs because of their reliance on high resources and low learning capacity. Prompt tuning (PT) has recently emerged as an effective fine-tuning method for edge LLMs by only modifying a small portion of LLM parameters, but it suffers from user domain shifts, resulting in repetitive training and losing resource efficiency. Conventional techniques to address domain shift issues often involve complex neural networks and sophisticated training, which are incompatible for PT for edge LLMs. Therefore, an open research question is how to address domain shift issues for edge LLMs with limited resources. In this paper, we propose a prompt tuning framework for edge LLMs, exploiting the benefits offered by non-volatile computing-in-memory (NVCiM) architectures. We introduce a novel NVCiM-assisted PT framework, where we narrow down the core operations to matrix-matrix multiplication, which can then be accelerated by performing in-situ computation on NVCiM. To the best of our knowledge, this is the first work employing NVCiM to improve the edge LLM PT performance.

Figures

Figures reproduced from arXiv: 2411.08244 by the authors.

Figure 1
Figure 1. Edge LLM performance comparison for two LLMs (Gemma [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Resource by the storing virtual tokens and data moving. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of our proposed NVCiM-based Prompt Tuning framework (NVCiM-PT). By co-design, it can utilize different types of NVCiM to improve prompt-tuning-based LLM content generation. the typical PT pipeline from a one-for-all approach to search￾ing for OVTs, designed to effectively leverage NVCiM. NVCiM-PT is specifically tailored for resource-constrained environments where LLMs need to learn on edge devices, enablin… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Implementation of our scaled search algorithm on NVCiM [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Evaluation of our SSA on CPU, RRAM, and FeFET [PITH_FULL_IMAGE:figures/full_fig_p006_5.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. Tiny-Align: Bridging Automatic Speech Recognition and Large Language Model on the Edge

    cs.SD 2024-11 reject novelty 4.0 of 10

    Tiny-Align aligns ASR audio features with an LLM's text-embedding space via a trained projector, claiming 50x faster convergence and improved ROUGE scores for edge ASR-LLM personalization.

Reference graph

Works this paper leans on

36 extracted references · 23 canonical work pages · cited by 1 Pith paper

  1. [1]

    FL-NAS: Towards Fairness of NAS for Resource Constrained Devices via Large Language Models

    Ruiyang Qin, Yuting Hu, Zheyu Yan, Jinjun Xiong, Ahmed Abbasi, and Yiyu Shi. Fl-nas: Towards fairness of nas for resource constrained devices via large language models. arXiv preprint arXiv:2402.06696 , 2024

    work page Pith review arXiv 2024

  2. [2]

    Language models for online depression detection: A review and benchmark analysis on remote interviews

    Ruiyang Qin, Ryan Cook, Kai Yang, Ahmed Abbasi, David Dobolyi, Salman Seyedi, Emily Griner, Hyeokhyen Kwon, Robert Cotes, Zifan Jiang, et al. Language models for online depression detection: A review and benchmark analysis on remote interviews. ACM Transactions on Management Information Systems , 2024

    2024

  3. [3]

    When Automated Assessment Meets Automated Content Generation: Examining Text Quality in the Era of GPTs

    Marialena Bevilacqua, Kezia Oketch, Ruiyang Qin, Will Stamey, Xinyuan Zhang, Yi Gan, Kai Yang, and Ahmed Abbasi. When automated assessment meets automated content generation: Examining text quality in the era of gpts. arXiv preprint arXiv:2309.14488 , 2023

    work page Pith review arXiv 2023

  4. [4]

    Privacy issues in large language models: A survey, 2023

    Seth Neel and Peter Chang. Privacy issues in large language models: A survey, 2023

    work page 2023

  5. [5]

    Embracing large language models for medical appli- cations: Opportunities and challenges

    Karabacak et al. Embracing large language models for medical appli- cations: Opportunities and challenges. Cureus, May 2023

    work page 2023

  6. [6]

    Can large language models be good companions? an llm-based eyewear system with conversational common ground, 2023

    Xu et al. Can large language models be good companions? an llm-based eyewear system with conversational common ground, 2023

    work page 2023

  7. [7]

    Personal llm agents: Insights and survey about the capability, efficiency and security, 2024

    Li et al. Personal llm agents: Insights and survey about the capability, efficiency and security, 2024

    work page 2024

  8. [8]

    Robust implementation of retrieval-augmented generation on edge-based computing-in-memory architectures

    Ruiyang Qin, Zheyu Yan, Dewen Zeng, Zhenge Jia, Dancheng Liu, Jianbo Liu, Zhi Zheng, Ningyuan Cao, Kai Ni, Jinjun Xiong, et al. Robust implementation of retrieval-augmented generation on edge-based computing-in-memory architectures. arXiv preprint arXiv:2405.04700 , 2024

    arXiv 2024

Show all 36 references
  1. [9]

    Enabling on-device large language model personalization with self-supervised data selection and synthesis

    Ruiyang Qin, Jun Xia, Zhenge Jia, Meng Jiang, Ahmed Abbasi, Peipei Zhou, Jingtong Hu, and Yiyu Shi. Enabling on-device large language model personalization with self-supervised data selection and synthesis. arXiv preprint arXiv:2311.12275 , 2023

    2023 arXiv

  2. [10]

    Empirical guidelines for deploying llms onto resource-constrained edge devices

    Ruiyang Qin, Dancheng Liu, Zheyu Yan, Zhaoxuan Tan, Zixuan Pan, Zhenge Jia, Meng Jiang, Ahmed Abbasi, Jinjun Xiong, and Yiyu Shi. Empirical guidelines for deploying llms onto resource-constrained edge devices. arXiv preprint arXiv:2406.03777 , 2024

    2024 arXiv

  3. [11]

    Prefix-tuning: Optimizing continuous prompts for generation

    Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimizing continuous prompts for generation. arXiv preprint arXiv:2101.00190 , 2021

    2021 arXiv

  4. [12]

    Dept: Decomposed prompt tuning for parameter-efficient fine-tuning

    Zhengxiang Shi and Aldo Lipani. Dept: Decomposed prompt tuning for parameter-efficient fine-tuning. arXiv preprint arXiv:2309.05173, 2023

    2023 arXiv

  5. [13]

    Pi- whisper: An adaptive and incremental asr framework for diverse and evolving speaker characteristics

    Amir Nassereldine, Dancheng Liu, Chenhui Xu, and Jinjun Xiong. Pi- whisper: An adaptive and incremental asr framework for diverse and evolving speaker characteristics. arXiv preprint arXiv:2406.15668, 2024

    2024 arXiv

  6. [14]

    The power of scale for parameter-efficient prompt tuning

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

    2021 arXiv

  7. [15]

    P-tuning v2: Prompt tuning can be comparable to fine-tuning universally across scales and tasks

    Xiao Liu, Kaixuan Ji, Yicheng Fu, Weng Lam Tam, Zhengxiao Du, Zhilin Yang, and Jie Tang. P-tuning v2: Prompt tuning can be comparable to fine-tuning universally across scales and tasks. arXiv preprint arXiv:2110.07602, 2021

    2021 arXiv

  8. [16]

    Ferroelectric compute-in-memory annealer for combinatorial optimization problems

    Xunzhao Yin, Yu Qian, Alptekin Vardar, Marcel G ¨unther, Franz M ¨uller, Nellie Laleni, Zijian Zhao, Zhouhang Jiang, Zhiguo Shi, Yiyu Shi, et al. Ferroelectric compute-in-memory annealer for combinatorial optimization problems. Nature Communications, 15(1):2419, 2024

    2024

  9. [17]

    A crossbar array of magnetoresistive memory devices for in-memory computing

    Seungchul Jung, Hyungwoo Lee, Sungmeen Myung, Hyunsoo Kim, Seung Keun Yoon, Soon-Wan Kwon, Yongmin Ju, Minje Kim, Wooseok Yi, Shinhee Han, et al. A crossbar array of magnetoresistive memory devices for in-memory computing. Nature, 601(7892):211–216, 2022

    2022

  10. [18]

    A compute-in-memory chip based on resistive random-access memory

    Weier Wan, Rajkumar Kubendran, Clemens Schaefer, Sukru Burc Ery- ilmaz, Wenqiang Zhang, Dabin Wu, Stephen Deiss, Priyanka Raina, He Qian, Bin Gao, et al. A compute-in-memory chip based on resistive random-access memory. Nature, 608(7923):504–512, 2022

    2022

  11. [19]

    The future of ferroelectric field-effect transistor technology

    Asif Islam Khan, Ali Keshavarzi, and Suman Datta. The future of ferroelectric field-effect transistor technology. Nature Electronics , 3(10):588–597, 2020

    2020

  12. [20]

    Swim: Selective write-verify for computing-in-memory neural accelerators

    Yan et al. Swim: Selective write-verify for computing-in-memory neural accelerators. In 2022 59th ACM/IEEE Design Automation Conference (DAC). IEEE

    2022

  13. [21]

    Uncertainty modeling of emerging device based computing- in-memory neural accelerators with application to neural architecture search

    Yan et al. Uncertainty modeling of emerging device based computing- in-memory neural accelerators with application to neural architecture search. In 2021 26th Asia and South Pacific Design Automation Conference (ASP-DAC). IEEE, 2021

    2021

  14. [22]

    Signal and noise extraction from analog memory elements for neuromorphic computing

    Nanbo Gong, T Id ´e, S Kim, Irem Boybat, Abu Sebastian, Vijay Narayanan, and Takashi Ando. Signal and noise extraction from analog memory elements for neuromorphic computing. Nature communications, 9(1):2102, 2018

    2018

  15. [23]

    On the reliability of computing-in-memory accelerators for deep neural networks

    Zheyu Yan, Xiaobo Sharon Hu, and Yiyu Shi. On the reliability of computing-in-memory accelerators for deep neural networks. In System Dependability and Analytics: Approaching System Dependability from Data, System and Analytics Perspectives , pages 167–190. Springer, 2022

    2022

  16. [24]

    Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding

    Song Han, Huizi Mao, and William J Dally. Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding. arXiv preprint arXiv:1510.00149 , 2015

    2015 arXiv

  17. [25]

    Text similarity estimation based on word embeddings and matrix norms for targeted marketing

    Tim vor der Br ¨uck and Marc Pouly. Text similarity estimation based on word embeddings and matrix norms for targeted marketing. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, V...

    2019

  18. [26]

    From word embeddings to pre-trained language models: A state-of-the-art walkthrough

    Mourad Mars. From word embeddings to pre-trained language models: A state-of-the-art walkthrough. Applied Sciences, 12(17):8805, 2022

    2022

  19. [27]

    Parameter-efficient fine-tuning methods for pretrained lan- guage models: A critical review and assessment

    Lingling Xu, Haoran Xie, Si-Zhao Joe Qin, Xiaohui Tao, and Fu Lee Wang. Parameter-efficient fine-tuning methods for pretrained lan- guage models: A critical review and assessment. arXiv preprint arXiv:2312.12148, 2023

    2023 arXiv

  20. [28]

    Baseline needs more love: On simple word-embedding- based models and associated pooling mechanisms

    Dinghan Shen, Guoyin Wang, Wenlin Wang, Martin Renqiang Min, Qinliang Su, Yizhe Zhang, Chunyuan Li, Ricardo Henao, and Lawrence Carin. Baseline needs more love: On simple word-embedding- based models and associated pooling mechanisms. arXiv preprint arXiv:1805.09843, 2018

    2018 arXiv

  21. [29]

    Fully hardware-implemented memristor convolutional neural network

    Yao et al. Fully hardware-implemented memristor convolutional neural network. Nature, 577(7792):641–646, 2020

    2020

  22. [30]

    Architecture-circuit-technology co-optimization for resistive random access memory-based computation-in-memory chips

    Liu et al. Architecture-circuit-technology co-optimization for resistive random access memory-based computation-in-memory chips. Science China Information Sciences , 66(10):200408, 2023

    2023

  23. [31]

    Switching pathway-dependent strain-effects on the ferroelec- tric properties and structural deformations in orthorhombic hfo2

    Wei et al. Switching pathway-dependent strain-effects on the ferroelec- tric properties and structural deformations in orthorhombic hfo2. Journal of Applied Physics , 131(15), 2022

    2022

  24. [32]

    Rouge: A package for automatic evaluation of sum- maries

    Chin-Yew Lin. Rouge: A package for automatic evaluation of sum- maries. In Text summarization branches out , pages 74–81, 2004

    2004

  25. [33]

    Cxdnn: Hardware-software compensation methods for deep neural networks on resistive crossbar systems

    Shubham Jain and Anand Raghunathan. Cxdnn: Hardware-software compensation methods for deep neural networks on resistive crossbar systems. ACM Transactions on Embedded Computing Systems (TECS) , 18(6):1–23, 2019

    2019

  26. [34]

    Correctnet: Robustness enhancement of analog in- memory computing for neural networks by error suppression and com- pensation

    Amro Eldebiky, Grace Li Zhang, Georg B ¨ocherer, Bing Li, and Ulf Schlichtmann. Correctnet: Robustness enhancement of analog in- memory computing for neural networks by error suppression and com- pensation. In 2023 Design, Automation & Test in Europe Conference & Exhibition (D...

    2023

  27. [35]

    Learning binary codes for maximum inner product search

    Fumin Shen, Wei Liu, Shaoting Zhang, Yang Yang, and Heng Tao Shen. Learning binary codes for maximum inner product search. In Proceed- ings of the IEEE International Conference on Computer Vision , pages 4148–4156, 2015

    2015

  28. [36]

    Dnn+ neurosim v2

    Xiaochen Peng, Shanshi Huang, Hongwu Jiang, Anni Lu, and Shimeng Yu. Dnn+ neurosim v2. 0: An end-to-end benchmarking framework for compute-in-memory accelerators for on-chip training. IEEE Trans- actions on Computer-Aided Design of Integrated Circuits and Systems , 40(11):2306...

    2020

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.