Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Solo Connection: A Parameter Efficient Fine-Tuning Technique for Transformers

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A new parameter-efficient fine-tuning method, Solo Connection, adapts GPT-2 by adding trainable long skip connections between decoder blocks, beating LoRA on E2E generation with 59% fewer trainable parameters.

desk verdict Plausible incremental PEFT with an unverified headline comparison; worth reviewing but needs major revisions. read the letter →

arxiv 2507.14353 v1 pith:2FKXPMT5 submitted 2025-07-18 cs.LG cs.AIcs.CL

Harsh Nilesh Pathak , Randy Paffenroth This is my paper · ORCID
classification cs.LGcs.AIcs.CL
keywords parameter-efficientfine-tuningLoRAlongskipconnectionshomotopyGPT-2low-rankadaptationE2ENLGbenchmarktransformer
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 proposes Solo Connection, a parameter-efficient fine-tuning (PEFT) technique that adapts a frozen transformer by adding trainable skip connections between decoder blocks rather than modifying attention weight matrices. The central claim is that this inter-layer adaptation outperforms LoRA on the E2E natural language generation benchmark while using 59% fewer trainable parameters on GPT-2 Small (0.12M versus 0.29M) and more than 99% fewer than full fine-tuning. If true, the result suggests that large language models can be specialized to new tasks at a fraction of the adaptation cost, and that block-level representation adaptation is a viable alternative to intra-layer low-rank updates. The paper also argues that a homotopy-inspired scaling layer makes the adaptation smooth and stable.

What carries the argument

The central object is the Solo Connection module, a trainable block composed of a dropout layer, a shared low-rank encoder $f_{se}$, a task-specific encoding vector $f_{ev}$, a shared decoder $f_{sd}$, and a homotopy linear layer $f_h(z)=\lambda v \odot z + (1-\lambda)0$. The encoder and decoder are shared across every Solo Connection in the model, so adding more connections adds only the encoding vector and homotopy parameters; sparsity masks a fraction of the encoder and decoder weights. Placed between consecutive decoder blocks (every other block in the experiments), this module creates a long skip connection that carries an adapted representation forward. The homotopy layer, with λ initialized near 0 and bounded in [0,1], gradually mixes the zero vector with the adapted representation, which the authors argue stabilizes training.

What would settle it

Re-run the E2E experiment from scratch with the identical data split, tokenizer, batch size, and training steps for both Solo Connection and LoRA, using the same random seeds and computing budget; if Solo Connection's BLEU does not exceed LoRA's at the claimed parameter count, the paper's headline result fails.

Watch

Extended reading notes

Core claim

The paper establishes that Solo Connection—a module inserted between consecutive decoder blocks, with a shared low-rank, sparse encoder–decoder and a trainable homotopy scaling λ—can outperform LoRA on five E2E NLG metrics while reducing trainable parameters from 0.29M to 0.12M for GPT-2 Small (BLEU 67.64 vs 65.79) and from 0.35M to 0.26M for GPT-2 Medium (BLEU 67.7 vs 67.45). The authors attribute the gains to parameter sharing across decoder blocks and to the homotopy layer $f_h(z)=\lambda v \odot z + (1-\lambda)0$, which interpolates from a zero vector so the frozen model is perturbed gradually. Their ablations show the homotopy layer is essential: replacing it with a plain learned vector collapses training to near-zero BLEU.

Load-bearing premise

The performance comparison assumes that the reported LoRA and full fine-tuning baselines were produced under the same data split, tokenization, training budget, and evaluation as the Solo Connection runs, since some baseline numbers are taken from prior work rather than recomputed.

Editorial extensions

If this is right

  • A 12-layer GPT-2 can be adapted to a new generation task with roughly 0.12M trainable parameters, about one tenth of a single decoder block's weight matrix, making multi-task serving of many adapters on one GPU more practical.
  • Because the encoder and decoder are shared across all Solo Connections, the per-connection cost is mostly the task-specific encoding vector and homotopy parameters; the paper's span experiments show that sharing one connection across three decoder blocks retains most of the performance.
  • The homotopy scaling, with λ converging to values below 0.1, means the actual perturbation to the frozen model is small; this suggests that a lightweight corrective bias at block outputs may be sufficient for many NLG tasks.

Reading between the lines

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

  • The homotopy layer is essentially a continuation method applied to fine-tuning; one could test whether initializing λ at larger values or annealing it over a schedule changes the stability and final performance, as the paper only reports a fixed small initialization.
  • The span experiments suggest that skipping too many decoder blocks destroys performance; an adaptive placement strategy for Solo Connections in models with 48 or 128 blocks could outperform the fixed every-other-block pattern used here.
  • Because λ converges to a small value, the adapted representation is a small correction; this hints that PEFT gains may be largely a task-specific bias effect, which could be tested by comparing Solo Connection against a simple learned bias on each block output.
  • The paper reports metrics on E2E only; extending to SuperGLUE or instruction-tuning tasks would show whether the inter-layer adaptation benefits generalize beyond generation.
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 / 6 minor

Summary. The paper proposes Solo Connection, a parameter-efficient fine-tuning (PEFT) method for GPT-2 that adds trainable low-rank, sparse long skip connections between decoder blocks, with a shared encoder/decoder and a trainable scalar lambda that the authors call a homotopy layer. The central empirical claim is that Solo Connection outperforms LoRA on the E2E natural language generation benchmark while using up to 59% fewer trainable parameters than LoRA and more than 99% fewer than full fine-tuning. The paper reports results for GPT-2 Small and Medium, along with ablations over rank, sparsity, connection span, and trainability of the encoder/decoder.

Significance. If the empirical claims were fully supported, the paper would offer a simple and parameter-efficient alternative to LoRA, and the idea of adapting decoder-block representations through shared, sparse, inter-layer connections is worth investigating. The paper also includes useful ablation experiments and is explicit about some of its limitations. However, the central comparison is not currently established: the LoRA and full fine-tuning baselines are borrowed from prior work rather than rerun in the same harness, no error bars or seeds are reported, the hyperparameter tuning is asymmetric between methods, and the parameter-count formula in Appendix B does not reproduce the values in the main tables. The claim that Solo Connection outperforms full fine-tuning is also contradicted by the paper's own Table 1. I agree with the reader that there is no circularity in the method itself, but the load-bearing empirical evidence needs substantial repair.

major comments (3)
  1. [Section 3, Table 1, and Contribution 3] The claim that Solo Connection outperforms full fine-tuning is not supported by Table 1. For GPT-2 Medium, full fine-tuning achieves higher BLEU (68.2 vs 67.7), METEOR (46.2 vs 45.95), ROUGE (71.0 vs 69.13), and CIDEr (2.47 vs 2.36); Solo Connection is higher only on NIST (8.64 vs 8.62). Since Contribution 3 explicitly states that the method outperforms LoRA and full fine-tuning, the text and table need to be reconciled, or the claim must be removed.
  2. [Section 3] The headline comparison with LoRA and full fine-tuning is not apples-to-apples. The paper says it replicates the experimental setup of LoRA but then states that, due to resource limits, it uses previously reported results from Hu et al. Without running LoRA and full fine-tuning in the same harness, differences in data split, tokenization, decoding, metric computation, or training budget could account for the reported gaps. The comparison is further biased because LoRA baselines use their original hyperparameters while Solo Connection's learning rate and rank are tuned, and Tables 2 and 3 show many Solo Connection configurations from which the Table 1 row appears to be selected. No seeds or error bars are reported, so the +2.82 BLEU advantage on GPT-2 Small cannot be distinguished from noise. Please rerun all methods under identical conditions with multiple seeds, report variances, and either provide a working code link or a fully detailed configuration, or re-label the external numbers and soften the superiority claims.
  3. [Appendix B] The parameter-count formula does not reproduce the reported table values. For the paper's own example with d=1024, r=32, s=0.7, n=2, T=11, the formula gives 31,276.8, which matches the stated 31,276, but for GPT-2 Small with r=128, s=0.6, T=5, d=768 the formula gives about 83k parameters, while Table 2 reports 0.12M. Notably, the 0.12M value is close to what the formula gives with d=1024 and T=11, i.e., GPT-2 Medium dimensions, suggesting the counts may be computed for the wrong model. For GPT-2 Medium with r=512, s=0.7, T=11, d=1024, the formula gives about 331k, while Table 3 reports 0.26M. The architecture description says the encoder and decoder are shared across all Solo Connections, yet the formula includes per-layer terms r*T and d*T; please clarify which parameters are shared and which are per-connection, and reconcile the formula with the tables, since the paper's parameter-reduction claims are computed from these numbers.
minor comments (6)
  1. [Section 2.2 and Section 3] The text says Kaiming initialization ensures that there is no need to tune the learning rate for each rank, but Section 3 then says the authors modify the rank and tune the learning rate. Please clarify whether the learning rate is held fixed across ranks or tuned per configuration.
  2. [Section 3] The reproducibility statement, "All code and configurations are available on GitHub [Anonymous Link]," is a placeholder; the final version should provide a working link and should report the exact training epochs, learning rate schedule, warmup, and evaluation decoding settings.
  3. [Equation (3)] As written, f_h(z)=lambda v circled dot z + (1-lambda)0 simplifies to lambda v circled dot z, so the second term adds nothing. The paper also says lambda is bounded in [0,1] but does not specify the mechanism (e.g., clamping or a sigmoid parameterization) that enforces this bound during training, nor does it report the learned trajectory of lambda.
  4. [Section 2.1] The text says Solo Connections are applied to alternate decoder blocks starting from D2 to the final block; for GPT-2 Small with 12 decoder blocks this would yield 6 connections (D2, D4, D6, D8, D10, D12), not 5. If the final block is excluded, the text should say so explicitly, because the number of connections affects both the architecture and the parameter count.
  5. [Appendix C.1] Tables 2 and 3 do not clearly label which rows include sparsity; the row labels 'r=512' versus 'r=512, s=0.6' are easy to misread, and the text's grouping into 'with sparsity' and 'without sparsity' should be made explicit in the table formatting.
  6. [References] There are duplicate and incomplete references: 'Nilesh Pathak & Paffenroth, 2019' duplicates 'Pathak & Paffenroth, 2019', 'Strogatz' lacks year and publication details, and 'Anonymous' is not an acceptable citation in a final manuscript.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is an external benchmark comparison, and self-citations are motivational rather than load-bearing.

full rationale

The paper's load-bearing claim is the empirical comparison in Table 1: Solo Connection versus LoRA and full fine-tuning on the E2E benchmark. The Solo Connection parameters (encoder, decoder, and homotopy lambda/v) are learned from the task data, and no headline metric is reconstructed from a fitted constant or from the defining equations. Equations (1)-(3) define the architecture but do not contain the reported BLEU values, so there is no derivation chain in which the output equals the input. The LoRA and full-fine-tuning baselines are taken from Hu et al. (2021), which is a borrowed external result rather than a self-citation; any protocol mismatch would be a reproducibility or fairness concern, not circularity. The homotopy-related self-citations (Pathak & Paffenroth 2019, 2021; Hershey et al. 2024) are motivational and do not carry the central empirical claim. The algebraic form f_h(z)=lambda v*z+(1-lambda)0 reduces to lambda v*z, so the 'interpolation between zero vector and representation' language is loose, but this does not make any prediction equivalent to the paper's inputs. No circular step can be exhibited with a specific equation-to-equation reduction.

Assumptions & free parameters 4 free parameters · 3 assumptions · 1 invented entities

The method introduces trainable components whose sizes and initializations are chosen by hand, including rank, sparsity, lambda, and skip placement. It also assumes baseline comparability with prior reported numbers and assumes that the learned gate keeps the frozen model stable. No new physical entities are postulated; the Solo Connection module is the only new component.

free parameters (4)
  • Rank r = 128 (GPT-2 S), 512 (GPT-2 M) in headline runs
    Swept over 8, 32, 64, 128, 512, and 1024 in Appendix C.1. It controls the dimensionality of the shared encoder/decoder and affects both parameter count and all metrics.
  • Sparsity s = 0.6 (GPT-2 S), 0.7 (GPT-2 M) in headline runs
    Swept over 0.6 and 0.7. It masks a fraction of the shared encoder/decoder weights and is central to the parameter-reduction claim.
  • Homotopy scalar lambda = Initialized at 0.001; reported final values roughly 0 to 0.1
    Learned gate in Equation 3. Its initialization and [0,1] bound are chosen ad hoc, and the homotopy interpretation depends on this scalar.
  • Skip placement and span = Alternate blocks starting at D2; span 1 in headline runs, span 3 or 5 tested
    Where and how far the skip connection reaches is hand-chosen. Appendix C.2 shows large performance changes with span, so this choice shapes the headline result.
assumptions (3)
  • domain assumption Baseline numbers from Hu et al. (2021) are directly comparable to the new Solo Connection runs
    Section 3 says the setup replicates LoRA but also that reported results from Hu et al. (2021) are used. Any mismatch in data split, tokenization, training budget, or evaluation breaks the comparison.
  • ad hoc to paper Equation 3 with a trainable constant lambda implements a homotopy between zero and the task representation
    The paper calls lambda interpolation, but lambda is learned without a time schedule. This is an interpretive framing rather than a mathematical guarantee.
  • domain assumption Adding trainable residual paths around frozen decoder blocks does not destabilize the pretrained representation
    Equation 1 adds f_solo to frozen block outputs. The paper provides only E2E metrics as evidence and no analysis of representation drift. Appendix C.4 shows that removing the gate causes training collapse, so stability depends on this assumption.
invented entities (1)
  • Solo Connection module
    purpose: A trainable cross-block skip path with shared low-rank encoder/decoder, per-connection vectors, and a learned gate; adapts decoder representations without modifying pretrained weights.
    It is introduced by this paper and validated only on the E2E benchmark with point estimates. There is no independent formal or experimental support beyond the paper's own tables.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Solo Connection: A Parameter Efficient Fine-Tuning Technique for Transformers." pith.science (2026). https://pith.science/paper/2FKXPMT5

@misc{pith2026250714353,
  author       = {Pith},
  title        = {Pith review of: Solo Connection: A Parameter Efficient Fine-Tuning Technique for Transformers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2FKXPMT5}},
  note         = {Machine review of arXiv:2507.14353}
}
read the original abstract

Parameter efficient fine tuning (PEFT) is a versatile and extensible approach for adapting a Large Language Model (LLM) for newer tasks. One of the most prominent PEFT approaches, Low Rank Adaptation (LoRA), primarily focuses on adjusting the attention weight matrices within individual decoder blocks of a Generative Pre trained Transformer (GPT2). In contrast, we introduce Solo Connection a novel method that adapts the representation at the decoder-block level rather than modifying individual weight matrices. Not only does Solo Connection outperform LoRA on E2E natural language generation benchmarks, but it also reduces the number of trainable parameters by 59% relative to LoRA and by more than 99% compared to full fine-tuning of GPT2, an early version of Large Language Models (LLMs). Solo Connection is also motivated by homotopy theory: we introduce a trainable linear transformation that gradually interpolates between a zero vector and the task-specific representation, enabling smooth and stable adaptation over time. While skip connections in the original 12 layer GPT2 are typically confined to individual decoder blocks, subsequent GPT2 variants scale up to 48 layers, and even larger language models can include 128 or more decoder blocks. These expanded architectures underscore the need to revisit how skip connections are employed during fine-tuning. This paper focuses on long skip connections that link outputs of different decoder blocks, potentially enhancing the model's ability to adapt to new tasks while leveraging pre-trained knowledge.

Figures

Figures reproduced from arXiv: 2507.14353 by the authors.

Figure 1
Figure 1. Solo Connection compared with LoRA setup LoRA (top) is Intra-decoder connection: Fine-tuning large language models with LoRA involves adding trainable modules within each decoder layer, creating an intra-decoder connection. Solo Connection (bottom), on the other hand, introduces an inter-decoder connection where the shared encoder and decoder are connected across different decoder layers. This approach directly adap… view at source ↗
Figure 2
Figure 2. Solo Connection that transforms the representation from a previous decoder and adapts during fine-tuning for subsequent decoders. Here representation embedding of D6 is fine-tuned for D8. Here we also show the building block of the solo connections, which has encoder-decoder learnable weights along with a homotopy layer. fh(z) = λv ⊙ z + (1 − λ)0 (3) Here, λ (scalar) and v (vector) are trainable parameters, while z … view at source ↗

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. Principled Curriculum Learning using Parameter Continuation Methods

    cs.LG 2025-07 reject novelty 3.0 of 10

    The paper applies pseudo-arclength continuation, a classical numerical method, to neural network optimization, claiming improved generalization over ADAM on small MNIST tasks.

Reference graph

Works this paper leans on

31 extracted references · 18 canonical work pages · cited by 1 Pith paper

  1. [1]

    and Georg, K

    Allgower, E. and Georg, K. Introduction to Numerical Continuation Methods. Society for Industrial and Applied Mathematics, 2003. doi:10.1137/1.9780898719154. URL https://epubs.siam.org/doi/abs/10.1137/1.9780898719154

    doi:10.1137/1.9780898719154 2003

  2. [2]

    The equivalence of finite and infinite impulse iterative neural networks

    Anonymous. The equivalence of finite and infinite impulse iterative neural networks. in preparation for submission as an arXiv.org preprint, 2023

    work page 2023

  3. [3]

    B it F it: Simple parameter-efficient fine-tuning for transformer-based masked language-models

    Ben Zaken, E., Goldberg, Y., and Ravfogel, S. B it F it: Simple parameter-efficient fine-tuning for transformer-based masked language-models. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pp.\ 1--9, Dublin, Ireland, May 2022. Association for Computational Linguistics. doi:10.18653/v1/2...

    doi:10.18653/v1/2022.acl-short.1 2022

  4. [4]

    V., and Salakhutdinov, R

    Dai, Z., Yang, Z., Yang, Y., Carbonell, J., Le, Q. V., and Salakhutdinov, R. Transformer-xl: Attentive language models beyond a fixed-length context, 2019

    work page 2019

  5. [5]

    A., Oldeman, B., Paffenroth, R., Sandstede, B., Wang, X., Zhang, C., et al

    Doedel, E., Champneys, A., Dercole, F., Fairgrieve, T., Kuznetsov, Y. A., Oldeman, B., Paffenroth, R., Sandstede, B., Wang, X., Zhang, C., et al. Auto-07p: Continuation and bifurcation software for ordinary differential equations. 2007

    work page 2007

  6. [6]

    J., Bollt, E

    Gauthier, D. J., Bollt, E. M., Griffith, A., and Barbosa, W. A. S. Next generation reservoir computing. CoRR, abs/2106.07688, 2021. URL https://arxiv.org/abs/2106.07688

    arXiv 2021

  7. [7]

    Delving deep into rectifiers: Surpassing human-level performance on imagenet classification

    He, K., Zhang, X., Ren, S., and Sun, J. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. CoRR, abs/1502.01852, 2015. URL http://arxiv.org/abs/1502.01852

    arXiv 2015

  8. [8]

    Deep residual learning for image recognition

    He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2016

    work page 2016

Show all 31 references
  1. [9]

    Exploring neural network structure through iterative neural networks: Connections to dynamical systems

    Hershey, Q. Exploring neural network structure through iterative neural networks: Connections to dynamical systems. Master's thesis, Worcester Polytechnic Institute, 2022

    2022

  2. [10]

    Rethinking the relationship between recurrent and non-recurrent neural networks: A study in sparsity, 2024

    Hershey, Q., Paffenroth, R., Pathak, H., and Tavener, S. Rethinking the relationship between recurrent and non-recurrent neural networks: A study in sparsity, 2024. URL https://arxiv.org/abs/2404.00880

    2024 arXiv

  3. [11]

    Parameter-efficient transfer learning for NLP

    Houlsby, N., Giurgiu, A., Jastrzebski, S., Morrone, B., De Laroussilhe, Q., Gesmundo, A., Attariyan, M., and Gelly, S. Parameter-efficient transfer learning for NLP . In Chaudhuri, K. and Salakhutdinov, R. (eds.), Proceedings of the 36th International Conference on Machine Lea...

    2019

  4. [12]

    Parameter-efficient transfer learning for NLP

    Houlsby, N., Giurgiu, A., Jastrzebski, S., Morrone, B., de Laroussilhe, Q., Gesmundo, A., Attariyan, M., and Gelly, S. Parameter-efficient transfer learning for NLP . CoRR, abs/1902.00751, 2019 b . URL http://arxiv.org/abs/1902.00751

    1902 arXiv

  5. [13]

    J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., and Chen, W

    Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., and Chen, W. Lora: Low-rank adaptation of large language models, 2021

    2021

  6. [14]

    J., Blankevoort, T., and Asano, Y

    Kopiczko, D. J., Blankevoort, T., and Asano, Y. M. Ve RA : Vector-based random matrix adaptation. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=NjNfLdxr3A

    2024

  7. [15]

    Deep learning

    LeCun, Y., Bengio, Y., and Hinton, G. Deep learning. nature, 521 0 (7553): 0 436--444, 2015

    2015

  8. [16]

    and Paffenroth, R

    Nilesh Pathak, H. and Paffenroth, R. Parameter continuation methods for the optimization of deep neural networks. pp.\ 1637--1643, 2019. doi:10.1109/ICMLA.2019.00268

    2019

  9. [17]

    The E 2 E dataset: New challenges for end-to-end generation

    Novikova, J., Du s ek, O., and Rieser, V. The E 2 E dataset: New challenges for end-to-end generation. In Jokinen, K., Stede, M., DeVault, D., and Louis, A. (eds.), Proceedings of the 18th Annual SIG dial Meeting on Discourse and Dialogue , pp.\ 201--206, Saarbr \"u cken, Germ...

    2017 doi

  10. [18]

    Pathak, H. N. Parameter continuation with secant approximation for deep neural networks. PhD thesis, Master's Thesis at Worcester Polytechnic Institute, 2018

    2018

  11. [19]

    Pathak, H. N. and Paffenroth, R. Parameter continuation methods for the optimization of deep neural networks. In 2019 18th IEEE International Conference on Machine Learning And Applications (ICMLA), pp.\ 1637--1643. IEEE, 2019

    2019

  12. [20]

    Pathak, H. N. and Paffenroth, R. Principled curriculum learning using parameter continuation methods. 2021

    2021

  13. [21]

    N., Paffenroth, R., and Hershey, Q

    Pathak, H. N., Paffenroth, R., and Hershey, Q. Sequentia12d: Organizing center of skip connections for transformers. In 2023 International Conference on Machine Learning and Applications (ICMLA), pp.\ 362--368, 2023. doi:10.1109/ICMLA58977.2023.00057

    2023

  14. [22]

    Language models are unsupervised multitask learners

    Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., and Sutskever, I. Language models are unsupervised multitask learners. 2019

    2019

  15. [23]

    Tied-lora: Enhancing parameter efficiency of lora with weight tying, 2024

    Renduchintala, A., Konuk, T., and Kuchaiev, O. Tied-lora: Enhancing parameter efficiency of lora with weight tying, 2024

    2024

  16. [24]

    U-net: Convolutional networks for biomedical image segmentation, 2015

    Ronneberger, O., Fischer, P., and Brox, T. U-net: Convolutional networks for biomedical image segmentation, 2015

    2015

  17. [25]

    Rumelhart, D. E. and McClelland, J. L. Learning Internal Representations by Error Propagation, pp.\ 318--362. The MIT Press, 1987

    1987

  18. [26]

    Fundamentals of recurrent neural network ( RNN ) and long short-term memory ( LSTM ) network

    Sherstinsky, A. Fundamentals of recurrent neural network ( RNN ) and long short-term memory ( LSTM ) network. Physica D: Nonlinear Phenomena, 404: 0 132306, mar 2020. doi:10.1016/j.physd.2019.132306

    2020

  19. [27]

    Strogatz, S. H. Nonlinear dynamics and chaos: with applications to physics, biology, chemistry, and engineering (studies in nonlinearity), volume 1

  20. [28]

    Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bashlykov, N., Batra, S., Bhargava, P., Bhosale, S., Bikel, D., Blecher, L., Ferrer, C. C., Chen, M., Cucurull, G., Esiobu, D., Fernandes, J., Fu, J., Fu, W., Fuller, B., Gao, C., Goswami, V., Goyal, N....

    2023 arXiv

  21. [29]

    J., Tao, X., and Wang, F

    Xu, L., Xie, H., Qin, S.-Z. J., Tao, X., and Wang, F. L. Parameter-efficient fine-tuning methods for pretrained language models: A critical review and assessment, 2023

    2023

  22. [30]

    Adalora: Adaptive budget allocation for parameter-efficient fine-tuning, 2023

    Zhang, Q., Chen, M., Bukharin, A., Karampatziakis, N., He, P., Cheng, Y., Chen, W., and Zhao, T. Adalora: Adaptive budget allocation for parameter-efficient fine-tuning, 2023

    2023

  23. [31]

    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 6, 2026 · model on record in the stance chip above.