Neural Incompatibility: The Unbridgeable Gap of Cross-Scale Parametric Knowledge Transfer in Large Language Models

Jun Zhao; Kang Liu; Shizhu He; Yuqiao Tan

REVIEW 1 major objections 6 minor 64 references

Neural Incompatibility: The Unbridgeable Gap of Cross-Scale Parametric Knowledge Transfer in Large Language Models

T0 review · 1 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Copying weights between LLM sizes fails without alignment, and even aligned transfer stays unstable.

desk verdict Useful negative result on cross-scale parameter transfer, but the 'unbridgeable' conclusion is stronger than the evidence: only one assumed layer/neuron correspondence was tested, and the paper's own GSM8K result nearly closes the gap. read the letter →

arxiv 2505.14436 v1 pith:BW6OPDEF submitted 2025-05-20 cs.CL cs.AI

Yuqiao Tan , Shizhu He , Kang Liu , Jun Zhao This is my paper

classification cs.CLcs.AI

keywords parametricknowledgetransfercross-scaleLLMsneuralincompatibilityLoRAinitializationlocalizationmodelalignmenthypernetworkLlama-2

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 asks whether the knowledge stored in a larger language model's weights can be transplanted directly into a smaller model as a cheap alternative to distillation or fine-tuning. It argues that alignment of the two parametric spaces is a fundamental prerequisite, and that transfer between models of different scales remains unstable even when pre-alignment is attempted, because of what it calls Neural Incompatibility: the behavioral and weight-structure differences between models of different sizes. The paper introduces LaTen, a low-cost pre-alignment method that locates task-relevant neurons in the larger model and projects them into the smaller model's space with a small hypernetwork, but reports that results are not consistently stable across tasks. If the claim holds, weight-based cross-scale transfer is not a reliable shortcut, and any working parametric transfer must first solve the correspondence problem between the two models.

What carries the argument

The central machinery is the Locate-Then-Align (LaTen) pipeline for pre-aligned transfer, together with the diagnostic metrics used to expose incompatibility. LaTen first applies a static neuron-level attribution method to score each FFN and MHSA neuron in the larger model by how much removing it changes the predicted token's log-probability; it selects the top-ranked layers and the top neurons per layer, then sends the extracted delta parameters through a two-layer MLP hypernetwork trained on fewer than 100 examples with standard language-modeling loss, so the projected delta can be added directly to the smaller model's weights. The incompatibility diagnosis uses Centered Kernel Alignment (CKA) to compare layer representations between the two models and cosine similarity between LoRA parameters and the original or residual weights, contrasting PiSSA (which decomposes the target model's own weights) with SEEKING (which imports weights from the larger model).

What would settle it

Fit a learned permutation between the 13B and 7B layers, or a soft assignment between their neurons, before performing the same transfer; if transfer improves substantially or becomes stable, then Neural Incompatibility is largely an artifact of rank-based matching rather than a fundamental barrier.

Watch

Extended reading notes

Core claim

The paper's central claim is that alignment in parametric space is a necessary condition for transferring knowledge between differently sized language models through weights, and that this alignment is hard because the two models are 'neurally incompatible'. It redefines prior work such as SEEKING as Post-Align PKT, where extracted parameters initialize LoRA and alignment happens during later fine-tuning, and proposes Pre-Align PKT with LaTen, which uses neuron-level attribution to locate task-relevant FFN and MHSA parameters, an MLP hypernetwork to map them to the smaller model, and injection before any training on the task. The evidence shows that unaligned transfer severely damages the target model, that LoRA initialized from the larger model underperforms LoRA derived from the target model itself (PiSSA), that a task-specialized larger model transfers no better than a generic one, and that representation similarity measured by CKA between the 7B and 13B models is low, especially in attention modules. The paper concludes that Neural Incompatibility, expressed as weak ethological and parametric structural similarity between cross-scale LLMs, poses a fundamental challenge to achieving effective parametric knowledge transfer.

Load-bearing premise

The study assumes that the top-ranked layers and neurons of the larger model correspond to the smaller model's layers and neurons in the same order, so matching them by rank is meaningful; if the real correspondence is scrambled or distributed, the observed incompatibility may be partly an artifact of that assumption.

Editorial extensions

If this is right

Weight-based transfer from a larger to a smaller LLM cannot serve as a drop-in replacement for distillation or fine-tuning, because unaligned injection collapses task performance.
LoRA initialization derived from the target model's own weights (PiSSA) is a stronger baseline than initialization from a larger model's extracted parameters, so cross-model initialization must justify its added cost.
Pre-alignment with very few examples can yield task gains (for example, +4.40 on GSM8K) but is unstable, meaning practical use would require checkpoint selection and repeated runs.
A larger model that is specialized for a task transfers no better, and sometimes worse, than a generic larger model, so source capability alone does not predict transfer success.
Future parametric transfer methods must model the correspondence between source and target layers and neurons rather than assuming an index-based match.

Reading between the lines

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

If incompatibility is truly structural, then the search for universal neuron-to-neuron correspondences across model sizes may be fundamentally limited, and transfer may need to operate on distributed or functional units rather than individual neurons.
The failure of a stronger specialized source suggests the blocking factor is the source-target gap itself, not the strength of the extracted knowledge; a learned permutation or soft assignment between the two models' layers might reveal how much of the observed incompatibility is an artifact of rank-based matching.
LaTen's instability may reflect the optimization landscape of the hypernetwork rather than an intrinsic ceiling on cross-scale transfer, so a testable extension is to regularize the hypernetwork or search over different source layer orders.
The same diagnostic (CKA plus parametric cosine similarity) could be applied to same-scale models from different checkpoints to test whether incompatibility is specific to scale or a general property of transferring weights between any two independently trained models.

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.

Desk Editor's Note

read the letter

Short version: this is a useful negative result with a new framing, but the central claim is bigger than the experimental setup can support.

The paper does two things well. It shows, across several simple baselines, that directly injecting extracted parameters from a 13B model into a 7B model without alignment degrades performance — that is robust and convincing. And it organizes the space into Pre-Align and Post-Align PKT, which is a legitimate and clarifying way to think about it. LaTen, the hypernetwork alignment method, is a real new piece; on GSM8K it moves the 7B model from 16.07 to 20.47 against a 13B ceiling of 20.55, using only a few examples. The SEEKING vs PiSSA comparison — self-derived LoRA initialization beating cross-scale initialization — is also interesting evidence.

But the 'Neural Incompatibility' conclusion goes beyond the data. The experiments only test one assumed correspondence: the top-ranked layers and neurons of the 13B model, in index order, mapped directly to the 7B model. No learned mapping, no permutation test, no validation that this is the right alignment. Low transfer performance under that protocol conflates 'our assumed mapping is wrong' with 'no mapping exists.' And the paper's own best result is a counterexample to the word 'unbridgeable' — a gap of 0.08 on GSM8K is a bridge, not an unbridgeable chasm.

There are also scope issues. One model family, one size pair, no error bars on the main LaTen results. The 'stronger Ml' experiments use WizardCoder, CodeLlama, and WizardMath, which differ from Llama-2-13B in fine-tuning data as well as scale; the observed drop may be distribution shift, not incompatibility. And the 'first to comprehensively define' claim is a bit much given SEEKING's prior parametric transfer work.

The CKA and cosine similarity analyses are suggestive but correlational. The analogy to cross-species neuroscience is evocative, not explanatory.

I would send this to a serious referee. The setup is reproducible, the negative result is potentially effort-saving, and the PrePKT/LaTen piece deserves scrutiny. But the authors should be pushed to test the correspondence assumption directly, add uncertainty estimates, and soften the 'unbridgeable' claim to 'challenging under simple index-based alignment.' As is, it reads as a conditional negative result, not a law of nature.

Referee Report

1 major / 6 minor

Summary. The paper defines cross-scale Parametric Knowledge Transfer (PKT) as transferring knowledge from a larger LLM (M_l) to a smaller LLM (M_s) through weights, and argues that alignment in parameter space is a prerequisite for successful transfer. It distinguishes Post-Align PKT (PostPKT, e.g., SEEKING with LoRA initialization followed by fine-tuning) from a proposed Pre-Align PKT (PrePKT) paradigm, and introduces LaTen (Locate-Then-Align), which uses neuron-level attribution and a small hypernetwork to align extracted delta parameters with the target model using fewer than 100 examples. Experiments on MMLU, GSM8K, HumanEval, and MBPP with Llama-2-7B/13B-Chat show that unaligned transfer fails, that LaTen yields small and sometimes inconsistent gains, and that representations between the two scales have low CKA similarity. The paper attributes these failures to 'Neural Incompatibility,' defined as ethological and parametric structural differences between LLMs of different scales, and concludes that cross-scale parametric transfer faces fundamental, 'unbridgeable' obstacles.

Significance. If the central claim were established, the paper would provide a useful negative result for parametric knowledge transfer: cheap weight-based transfer from larger to smaller LLMs would be fundamentally limited, and the community would need to focus on learned alignment or distillation. The paper also makes a concrete proposal (LaTen) and releases code, and its controlled demonstration that unaligned injection degrades performance is a clean and reproducible observation. However, the significance is currently limited because the 'unbridgeable gap' claim rests on experiments that all use one fixed index-based correspondence between source and target layers/neurons, and because the reported gains for LaTen are small, partly within likely noise, and not reported with error bars for the PrePKT results. The stronger claims therefore need additional controls before they can support the paper's title-level conclusion.

major comments (1)

[Section 4.2 and Section 5.3] Section 5.3's CKA analysis is presented as evidence of 'ethological similarity' between scales, but the CKA values are computed between layer representations at the same layer indices (Figure 3). If the true correspondence between layers is nonlinear or non-monotonic, the low CKA could be an artifact of the index-aligned comparison. The paper should either show that the low-similarity conclusion is robust under learned feature-space alignments (e.g., using a fitted linear map before computing CKA) or state explicitly that the CKA result is conditional on index-based layer correspondence.

minor comments (6)

[Equations (1)-(2)] There are typos in the variable names: 'Delta Theta^T_algin' and 'Delta Theta^T_extarct' should be 'align' and 'extract'. Please correct these throughout.
[Appendix C.1] The heading 'Detailed Experimens' should be 'Detailed Experiments'.
[Section 1 (Contributions)] The word 'promissing' should be 'promising' in the third bullet point.
[Section 5.3] The term 'Neuron Incompatibility' is used in the text ('We call this Neuron Incompatibility') while the abstract and title use 'Neural Incompatibility.' Please make the terminology consistent.
[Table 1] The label 'Post-Align on D_train (=1000)' is misleading for MBPP, since Table 4 reports a training size of 300 for MBPP and the text in Section 5.1 says the training set is 1000 'except MBPP.' Please use the actual sizes in the table or in a footnote.
[Section 5.1 and Table 5] The alignment set sizes (32, 64, 48, 128) and the number of steps (2, 4, 3, 8) vary considerably across benchmarks. Please discuss how these were chosen and whether the results are sensitive to these choices; currently the paper gives no sensitivity analysis.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: alignment and LaTen are empirically controlled; Neural Incompatibility is an underdetermined explanation, not a built-in reduction.

full rationale

The paper's central derivation chain is not circular. The claim that alignment is a prerequisite is supported by controlled unaligned baselines (top/bottom/random layer selection, PCA/whitening/embedding transforms) that all degrade performance, while LaTen performs pre-alignment with a hypernetwork trained on under 100 examples and is compared against random-initialization LoRA and distillation baselines at the same data budget. The Neural Incompatibility explanation is inferred from independent CKA and cosine-similarity measurements that do not encode the transfer outcomes, so it is not a fitted parameter renamed as a prediction. The self-citations present (DyPRAG, Tan et al. 2025; whitening, Liao et al. 2024) are contextual related-work mentions and are not load-bearing for the central claims. The main substantive weakness is underdetermination, not circularity: the layer/neuron correspondence is assumed by index (top-Ls layers in order, top-c neurons by attribution), so failed transfers could partly reflect a wrong correspondence rather than intrinsic incompatibility, and the stronger-M_l comparisons use differently fine-tuned checkpoints that introduce distribution shift. These are confounds for the causal interpretation of 'Neural Incompatibility', and the paper's own GSM8K result (LaTen 20.47 vs 13B 20.55) shows near-total gap transfer on one benchmark, but none of this constitutes a reduction of a predicted result to its own inputs by construction.

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

The paper relies on several domain assumptions from the interpretability literature (knowledge neurons, index-based layer correspondence) and introduces one new conceptual entity (Neural Incompatibility) that is not independently falsifiable. The free parameters are hyperparameters of the proposed method.

free parameters (3)

number of alignment steps per dataset = 2 (MMLU), 4 (GSM8K), 3 (HumanEval), 8 (MBPP)
Chosen by hand per benchmark; the paper notes that identifying the best checkpoint requires multiple experiments, so this is a tuned hyperparameter affecting the reported gains.
neuron transfer ratio = 10% of neurons per layer
Set to 10% without sensitivity analysis; affects how much of the source model's knowledge is transferred.
hypernetwork training hyperparameters = lr=1e-5, weight_decay=0.05, P=16, MSE-to-zero constraint
Standard hyperparameters but chosen without reported sensitivity analysis; the MSE-to-zero constraint is an ad hoc regularizer.

assumptions (3)

domain assumption Knowledge neurons (FFN subvalues and attention subvalues) are meaningful units for knowledge transfer.
Adopted from Geva et al. (2020) and Yu and Ananiadou (2024); if neurons do not carry task knowledge, the extraction step in Section 4.2 is meaningless.
domain assumption The top-Ls layers from the source, in order, correspond to the target layers.
Section 4.2 assumes an index-based correspondence between the 13B and 7B models without evidence; this is load-bearing for extraction and injection.
domain assumption CKA similarity between model representations is a valid proxy for 'ethological similarity' and transferability.
Section 5.3 uses low CKA to explain transfer failure, but no baseline (e.g., same-scale models or random weights) is provided to calibrate what 'low' means.

invented entities (1)

Neural Incompatibility
purpose: To conceptualize the observed difficulty of cross-scale parameter transfer as a fundamental structural difference between models of different scales.
The term is a label for the observed failure, not an independently testable mechanism. The paper does not predict new observable phenomena beyond the experiments used to infer it.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Neural Incompatibility: The Unbridgeable Gap of Cross-Scale Parametric Knowledge Transfer in Large Language Models." pith.science (2026). https://pith.science/paper/BW6OPDEF

@misc{pith2026250514436,
  author       = {Pith},
  title        = {Pith review of: Neural Incompatibility: The Unbridgeable Gap of Cross-Scale Parametric Knowledge Transfer in Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BW6OPDEF}},
  note         = {Machine review of arXiv:2505.14436}
}

abstract

Large Language Models (LLMs) offer a transparent brain with accessible parameters that encode extensive knowledge, which can be analyzed, located and transferred. Consequently, a key research challenge is to transcend traditional knowledge transfer paradigms rooted in symbolic language and achieve genuine Parametric Knowledge Transfer (PKT). Significantly, exploring effective methods for transferring knowledge across LLMs of different scales through parameters presents an intriguing and valuable research direction. In this paper, we first demonstrate $\textbf{Alignment}$ in parametric space is the fundamental prerequisite to achieve successful cross-scale PKT. We redefine the previously explored knowledge transfer as Post-Align PKT (PostPKT), which utilizes extracted parameters for LoRA initialization and requires subsequent fine-tune for alignment. Hence, to reduce cost for further fine-tuning, we introduce a novel Pre-Align PKT (PrePKT) paradigm and propose a solution called $\textbf{LaTen}$ ($\textbf{L}$oc$\textbf{a}$te-$\textbf{T}$h$\textbf{e}$n-Alig$\textbf{n}$) that aligns the parametric spaces of LLMs across scales only using several training steps without following training. Comprehensive experiments on four benchmarks demonstrate that both PostPKT and PrePKT face challenges in achieving consistently stable transfer. Through in-depth analysis, we identify $\textbf{Neural Incompatibility}$ as the ethological and parametric structural differences between LLMs of varying scales, presenting fundamental challenges to achieving effective PKT. These findings provide fresh insights into the parametric architectures of LLMs and highlight promising directions for future research on efficient PKT. Our code is available at https://github.com/Trae1ounG/Neural_Incompatibility.

Figures

Figures reproduced from arXiv: 2505.14436 by the authors.

**Figure 2.** Performance of different baseline methods in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 4.** Parametric Similarity Comparison Results [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

Figures from the paper (5 more)

**Figure 5.** Figure 5: Interpretable neuron location in GSM8K task. [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

**Figure 6.** Figure 6: Results for Parametric Similarity Comparison [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]

**Figure 7.** Figure 7: Delta parameter ranges between Llama-2-7b and Llama-2-7b-Chat [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]

**Figure 8.** Figure 8: Delta parameter ranges between Llama-2-13b and Llama-2-13b-Chat [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]

**Figure 9.** Figure 9: Delta parameter ranges from Llama-2-13b-Chat in GSM8K using S [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

64 extracted references · 14 canonical work pages

[1]

Herv \'e Abdi and Lynne J Williams. 2010. Principal component analysis. Wiley interdisciplinary reviews: computational statistics, 2(4):433--459

2010
[2]

Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

arXiv 2023
[3]

Rishabh Agarwal, Nino Vieillard, Yongchao Zhou, Piotr Stanczyk, Sabela Ramos Garea, Matthieu Geist, and Olivier Bachem. 2024. On-policy distillation of language models: Learning from self-generated mistakes. In The Twelfth International Conference on Learning Representations

2024
[4]

Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, et al. 2021. Program synthesis with large language models. arXiv preprint arXiv:2108.07732

arXiv 2021
[5]

Tian Bowen, Lai Songning, Wu Jiemin, Shuai Zhihao, Ge Shiming, and Yue Yutao. 2024. Beyond task vectors: Selective task arithmetic based on importance metrics. arXiv preprint arXiv:2411.16139

arXiv 2024
[6]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877--1901

2020
[7]

Lihu Chen, Adam Dejl, and Francesca Toni. 2024 a . Analyzing key neurons in large language models. arXiv preprint arXiv:2406.10868

work page Pith review arXiv 2024
[8]

Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde De Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. 2021. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374

arXiv 2021

Show all 64 references

[9]

Wei Chen, Zhen Huang, Liang Xie, Binbin Lin, Houqiang Li, Le Lu, Xinmei Tian, Deng Cai, Yonggang Zhang, Wenxiao Wan, et al. 2024 b . From yes-men to truth-tellers: Addressing sycophancy in large language models with pinpoint tuning. arXiv preprint arXiv:2409.01658

2024 arXiv
[10]

Yuheng Chen, Pengfei Cao, Yubo Chen, Kang Liu, and Jun Zhao. 2024 c . Journey to the center of the knowledge neurons: Discoveries of language-independent knowledge neurons and degenerate knowledge neurons. In Proceedings of the AAAI Conference on Artificial Intelligence, volum...

2024
[11]

Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, et al. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168

2021 arXiv
[12]

Damai Dai, Li Dong, Yaru Hao, Zhifang Sui, Baobao Chang, and Furu Wei. 2021. Knowledge neurons in pretrained transformers. arXiv preprint arXiv:2104.08696

2021 arXiv
[13]

Nicola De Cao, Wilker Aziz, and Ivan Titov. 2021. Editing factual knowledge in language models. arXiv preprint arXiv:2104.08164

2021 arXiv
[14]

Jacob Devlin. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805

2018 arXiv
[15]

Alexey Dosovitskiy. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

2020 arXiv
[16]

Nelson Elhage, Neel Nanda, Catherine Olsson, Tom Henighan, Nicholas Joseph, Ben Mann, Amanda Askell, Yuntao Bai, Anna Chen, Tom Conerly, Nova DasSarma, Dawn Drain, Deep Ganguli, Zac Hatfield-Dodds, Danny Hernandez, Andy Jones, Jackson Kernion, Liane Lovitt, Kamal Ndousse, Dari...

2021
[17]

Mor Geva, Jasmijn Bastings, Katja Filippova, and Amir Globerson. 2023. Dissecting recall of factual associations in auto-regressive language models. arXiv preprint arXiv:2304.14767

2023 arXiv
[18]

Mor Geva, Roei Schuster, Jonathan Berant, and Omer Levy. 2020. Transformer feed-forward layers are key-value memories. arXiv preprint arXiv:2012.14913

2020 arXiv
[19]

Gene H Golub and Christian Reinsch. 1971. Singular value decomposition and least squares solutions. In Handbook for Automatic Computation: Volume II: Linear Algebra, pages 134--151. Springer

1971
[20]

Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874

2021 arXiv
[21]

Geoffrey Hinton. 2015. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531

2015 arXiv
[22]

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

2021 arXiv
[23]

Shih-Cheng Huang, Pin-Zu Li, Yu-Chi Hsu, Kuang-Ming Chen, Yu Tung Lin, Shih-Kai Hsiao, Richard Tsai, and Hung-Yi Lee. 2024. Chat vector: A simple approach to equip llms with instruction following and model alignment in new languages. In Proceedings of the 62nd Annual Meeting o...

2024
[24]

Gabriel Ilharco, Marco Tulio Ribeiro, Mitchell Wortsman, Suchin Gururangan, Ludwig Schmidt, Hannaneh Hajishirzi, and Ali Farhadi. 2022. Editing models with task arithmetic. arXiv preprint arXiv:2212.04089

2022 arXiv
[25]

Xisen Jin, Xiang Ren, Daniel Preotiuc-Pietro, and Pengxiang Cheng. 2022. Dataless knowledge fusion by merging weights of language models. arXiv preprint arXiv:2212.09849

2022 arXiv
[26]

Yoon Kim and Alexander M Rush. 2016. Sequence-level knowledge distillation. In Proceedings of the 2016 conference on empirical methods in natural language processing, pages 1317--1327

2016
[27]

Simon Kornblith, Mohammad Norouzi, Honglak Lee, and Geoffrey Hinton. 2019. Similarity of neural network representations revisited. In International conference on machine learning, pages 3519--3529. PMLR

2019
[28]

Gonzalez, Hao Zhang, and Ion Stoica

Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. Efficient memory management for large language model serving with pagedattention. In Proceedings of the ACM SIGOPS 29th Symposium on Operating S...

2023
[29]

Bohan Li, Hao Zhou, Junxian He, Mingxuan Wang, Yiming Yang, and Lei Li. 2020. On the sentence embeddings from pre-trained language models. arXiv preprint arXiv:2011.05864

2020 arXiv
[30]

Huanxuan Liao, Yao Xu, Shizhu He, Yuanzhe Zhang, Yanchao Hao, Shengping Liu, Kang Liu, and Jun Zhao. 2024. From instance training to instruction learning: Task adapters generation from instructions. arXiv preprint arXiv:2406.12382

2024 arXiv
[31]

Alisa Liu, Xiaochuang Han, Yizhong Wang, Yulia Tsvetkov, Yejin Choi, and Noah A Smith. 2024. Tuning language models by proxy. arXiv preprint arXiv:2401.08565

2024 arXiv
[32]

Xiqian Lu, Zhaoqi Hu, Yumeng Xin, Tianshu Yang, Ying Wang, Peng Zhang, Ning Liu, and Yi Jiang. 2024. Detecting biological motion signals in human and monkey superior colliculus: a subcortical-cortical pathway for biological motion perception. Nature Communications, 15(1):9606

2024
[33]

Haipeng Luo, Qingfeng Sun, Can Xu, Pu Zhao, Jianguang Lou, Chongyang Tao, Xiubo Geng, Qingwei Lin, Shifeng Chen, and Dongmei Zhang. 2023 a . Wizardmath: Empowering mathematical reasoning for large language models via reinforced evol-instruct. arXiv preprint arXiv:2308.09583

2023 arXiv
[34]

Ziyang Luo, Can Xu, Pu Zhao, Qingfeng Sun, Xiubo Geng, Wenxiang Hu, Chongyang Tao, Jing Ma, Qingwei Lin, and Daxin Jiang. 2023 b . Wizardcoder: Empowering code large language models with evol-instruct. arXiv preprint arXiv:2306.08568

2023 arXiv
[35]

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

2022
[36]

Fanxu Meng, Zhaohui Wang, and Muhan Zhang. 2024. Pissa: Principal singular values and singular vectors adaptation of large language models. Advances in Neural Information Processing Systems, 37:121038--121072

2024
[37]

Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov. 2022 a . Locating and editing factual associations in gpt. Advances in Neural Information Processing Systems, 35:17359--17372

2022
[38]

Kevin Meng, Arnab Sen Sharma, Alex Andonian, Yonatan Belinkov, and David Bau. 2022 b . Mass-editing memory in a transformer. arXiv preprint arXiv:2210.07229

2022 arXiv
[39]

Eric Mitchell, Charles Lin, Antoine Bosselut, Chelsea Finn, and Christopher D Manning. 2021. Fast model editing at scale. arXiv preprint arXiv:2110.11309

2021 arXiv
[40]

Michael C Mozer and Paul Smolensky. 1988. Skeletonization: A technique for trimming the fat from a network via relevance assessment. Advances in neural information processing systems, 1

1988
[41]

Jingcheng Niu, Andrew Liu, Zining Zhu, and Gerald Penn. 2024. What does the knowledge neuron thesis have to do with knowledge? arXiv preprint arXiv:2405.02421

2024 arXiv
[42]

nostalgebraist. 2020. https://www.lesswrong.com/posts/AcKRB8wDpdaN6v6ru/interpreting-gpt-the-logit-lens interpreting gpt: the logit lens . LessWrong

2020
[43]

Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 3...

2022
[44]

Baptiste Roziere, Jonas Gehring, Fabian Gloeckle, Sten Sootla, Itai Gat, Xiaoqing Ellen Tan, Yossi Adi, Jingyu Liu, Romain Sauvestre, Tal Remez, et al. 2023. Code llama: Open foundation models for code. arXiv preprint arXiv:2308.12950

2023 arXiv
[45]

Ran Song, Shizhu He, Shuting Jiang, Yantuan Xian, Shengxiang Gao, Kang Liu, and Zhengtao Yu. 2024. Does large language model contain task-specific neurons? In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 7101--7113

2024
[46]

Jianlin Su. 2021. https://spaces.ac.cn/archives/8069 You probably don't need bert-flow: A linear transformation comparable to bert-flow

2021
[47]

Sainbayar Sukhbaatar, Jason Weston, Rob Fergus, et al. 2015. End-to-end memory networks. Advances in neural information processing systems, 28

2015
[48]

Mukund Sundararajan, Ankur Taly, and Qiqi Yan. 2017. Axiomatic attribution for deep networks. In International conference on machine learning, pages 3319--3328. PMLR

2017
[49]

Yuqiao Tan, Shizhu He, Huanxuan Liao, Jun Zhao, and Kang Liu. 2025. Dynamic parametric retrieval augmented generation for test-time knowledge enhancement. arXiv preprint arXiv:2503.23895

2025 arXiv
[50]

Tianyi Tang, Wenyang Luo, Haoyang Huang, Dongdong Zhang, Xiaolei Wang, Xin Zhao, Furu Wei, and Ji-Rong Wen. 2024. Language-specific neurons: The key to multilingual capabilities in large language models. arXiv preprint arXiv:2402.16438

2024 arXiv
[51]

Keyu Tian, Yi Jiang, Zehuan Yuan, Bingyue Peng, and Liwei Wang. 2024. Visual autoregressive modeling: Scalable image generation via next-scale prediction. arXiv preprint arXiv:2404.02905

2024 arXiv
[52]

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

2023 arXiv
[53]

A Vaswani. 2017. Attention is all you need. Advances in Neural Information Processing Systems

2017
[54]

Jesse Vig, Sebastian Gehrmann, Yonatan Belinkov, Sharon Qian, Daniel Nevo, Yaron Singer, and Stuart Shieber. 2020. Investigating gender bias in language models using causal mediation analysis. Advances in neural information processing systems, 33:12388--12401

2020
[55]

Kevin Wang, Alexandre Variengien, Arthur Conmy, Buck Shlegeris, and Jacob Steinhardt. 2022. Interpretability in the wild: a circuit for indirect object identification in gpt-2 small. arXiv preprint arXiv:2211.00593

2022 arXiv
[56]

Yufan Wang, Luqi Cheng, Deying Li, Yuheng Lu, Changshuo Wang, Yaping Wang, Chaohong Gao, Haiyan Wang, Camilla T Erichsen, Wim Vanduffel, et al. 2025. The chimpanzee brainnetome atlas reveals distinct connectivity and gene expression profiles relative to humans. The Innovation, 6(2)

2025
[57]

Mitchell Wortsman, Gabriel Ilharco, Samir Ya Gadre, Rebecca Roelofs, Raphael Gontijo-Lopes, Ari S Morcos, Hongseok Namkoong, Ali Farhadi, Yair Carmon, Simon Kornblith, et al. 2022. Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasin...

2022
[58]

Jiayi Wu, Hao Sun, Hengyi Cai, Lixin Su, Shuaiqiang Wang, Dawei Yin, Xiang Li, and Ming Gao. 2024. Cross-model control: Improving multiple large language models in one-time training. arXiv preprint arXiv:2410.17599

2024 arXiv
[59]

Le Yu, Bowen Yu, Haiyang Yu, Fei Huang, and Yongbin Li. 2024. Language models are super mario: Absorbing abilities from homologous models as a free lunch. In Forty-first International Conference on Machine Learning

2024
[60]

Zeping Yu and Sophia Ananiadou. 2024. Neuron-level knowledge attribution in large language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 3267--3280

2024
[61]

Jinghan Zhang, Junteng Liu, Junxian He, et al. 2023. Composing parameter-efficient modules with arithmetic operation. Advances in Neural Information Processing Systems, 36:12589--12610

2023
[62]

Ming Zhong, Chenxin An, Weizhu Chen, Jiawei Han, and Pengcheng He. 2023. Seeking neural nuggets: Knowledge transfer in large language models from a parametric perspective. arXiv preprint arXiv:2310.11451

2023 arXiv
[63]

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...
[64]

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.

[1] [1]

Herv \'e Abdi and Lynne J Williams. 2010. Principal component analysis. Wiley interdisciplinary reviews: computational statistics, 2(4):433--459

2010

[2] [2]

Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

arXiv 2023

[3] [3]

Rishabh Agarwal, Nino Vieillard, Yongchao Zhou, Piotr Stanczyk, Sabela Ramos Garea, Matthieu Geist, and Olivier Bachem. 2024. On-policy distillation of language models: Learning from self-generated mistakes. In The Twelfth International Conference on Learning Representations

2024

[4] [4]

Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, et al. 2021. Program synthesis with large language models. arXiv preprint arXiv:2108.07732

arXiv 2021

[5] [5]

Tian Bowen, Lai Songning, Wu Jiemin, Shuai Zhihao, Ge Shiming, and Yue Yutao. 2024. Beyond task vectors: Selective task arithmetic based on importance metrics. arXiv preprint arXiv:2411.16139

arXiv 2024

[6] [6]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877--1901

2020

[7] [7]

Lihu Chen, Adam Dejl, and Francesca Toni. 2024 a . Analyzing key neurons in large language models. arXiv preprint arXiv:2406.10868

work page Pith review arXiv 2024

[8] [8]

Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde De Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. 2021. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374

arXiv 2021