Pith. sign in

REVIEW 4 major objections 5 minor 52 references

QueueEDIT: Structural Self-Correction for Sequential Model Editing in LLMs

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Sequential LLM editing stays accurate over 1,000 updates by queueing and realigning related knowledge parameters.

desk verdict QueueEDIT has a novel and empirically promising mechanism for sequential model editing, but the paper's central update rule is under-specified to the point of being unimplementable as written. read the letter →

arxiv 2506.17864 v1 pith:HDK6DPVI submitted 2025-06-22 cs.CL

Taolin Zhang , Haidong Kang , Dongyang Li , Qizhou Chen , Chengyu Wang Xiaofeng He , Richang Hong This is my paper · ORCID
classification cs.CL
keywords sequentialmodeleditingknowledgelargelanguagemodelsqueue-basedself-correctionstructuralmappinglossparameteralignmentFIFOweightqueuefactualreliabilitygeneralitylocality
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

Sequential model editing aims to keep correcting a large language model's factual errors over many updates, but earlier edits tend to interfere with later ones and degrade general abilities. This paper claims that the interference can be managed by treating edited knowledge as a queue: each edit's located parameter matrix is enqueued, the most similar older matrices are realigned to the current edit, and irrelevant matrices are frozen until dequeued. On ZSRE, CounterFact, and RIPE at 1,000 edits on LLaMA3-8B, the method reports average scores of 69.1, 61.8, and 61.2, above the strongest baseline's 65.1, 55.3, and 55.6, while staying competitive in single-turn editing. The larger point is that a small structural bookkeeping mechanism can preserve both edit reliability and general language ability across long edit sequences.

What carries the argument

The carrying mechanism is a weight queue of located projection matrices $W_{\text{proj}}$ plus a translation-based structural loss. The loss is $L_{\text{st}} = \|k_s^* W_{\text{proj}} + h_r - v^*\|^2$, where $k_s^*$ is the subject-token key vector, $h_r$ is a relation embedding, and $v^*$ is the target object value; it forces the update to behave like $(s + r) \to o$. The queue then selects, by Euclidean distance in parameter space, the top-$K$ older matrices that are semantically close to the current edit and realigns them with $\Delta W = v_t^* \oplus h_r^i$ before concatenating and activating; freezing everything else preserves general capabilities.

What would settle it

Run a controlled sequential-editing experiment on 1,000 facts in which parameter-space distance and semantic relatedness are deliberately decoupled — for example, unrelated facts whose located matrices happen to be close, and related facts on the same entity whose matrices are far apart. If the queue's distance-based top-K selection does not outperform random selection or selection by true relation overlap, the self-correction signal is not semantic and the reported gains would not transfer outside the benchmark distributions.

Watch

Extended reading notes

Core claim

The paper's central claim is that sequential model editing degrades because each edit rewrites the same projection matrix independently, ignoring semantic dependencies among facts. QueueEDIT stores the projection matrix located for each edited triple in a FIFO weight queue. At each new edit it computes Euclidean distances between the current matrix and queued matrices, selects the top-K below a threshold, and applies a self-correction update that links the new object to the old relation; the queue head is dequeued when its distance falls below another threshold. A separate structural editing loss maps the triple's subject, relation, and object onto different parts of the FFN layer, and the paper reports that this combination beats modifying-parameter, extra-parameter, and retrieval baselines while keeping general NLP scores close to the unedited model.

Load-bearing premise

The load-bearing premise is that the Euclidean distance between two located projection matrices is a faithful stand-in for whether two edits concern related facts, so the top-K closest matrices are the ones needing realignment; the paper supports this with illustrative cases rather than quantitative correlation.

Editorial extensions

If this is right

  • At 1,000 edits on LLaMA3-8B, average scores rise to 69.1, 61.8, and 61.2 on ZSRE, CounterFact, and RIPE, versus 65.1, 55.3, and 55.6 for the strongest baseline DAFNet.
  • The same queue mechanism remains competitive at single-turn editing, so adopting it does not sacrifice one-shot edit quality for long-sequence stability.
  • General NLP capability, measured on public QA benchmarks, stays near the unedited model even after 1,000 edits, while baseline methods decline.
  • Using a larger backbone, Qwen2.5-14B, raises performance further, consistent with the claim that bigger models hold more internal knowledge for the queue to align.
  • Queue length trades memory against accuracy: 30% of 1,000 edits works best, while longer queues add semantic noise.

Reading between the lines

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

  • Editorial inference: if Euclidean proximity among located projection matrices really tracks factual relatedness, then any locate-then-edit method could be retrofitted with the queue by storing its own located matrices, giving a generic stabilizer for sequential editing.
  • Editorial inference: the queue's distance threshold could be made adaptive per relation or per layer; the paper fixes one threshold for all edits, so a testable extension is to tune $\eta_{\text{que}}$ against relation diversity.
  • Editorial inference: the evaluation is built from knowledge-graph triples, so a stress test would be to run the same queue on edit sequences with deliberately entangled subjects where the relevant dependency is not nearest in parameter space; the paper's case studies only show same-entity or same-relation neighbors.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. QueueEDIT proposes a sequential model editing (SME) framework for LLMs. It first introduces a structural mapping editing loss that maps a knowledge triple (s, r, o) to designated positions in an FFN layer: the subject is mapped to a key vector k_s^*, the relation to a relation embedding h_r, and the object to a target vector v_*; a translation-style loss (Eq. 11) then updates the projection matrix W_proj^{l0}. It then maintains a FIFO queue of located parameter matrices W_proj for past edits. At each new edit, the method computes Euclidean distances between the current matrix and queued matrices (Eq. 12), selects the top-K closest ones below a threshold, and applies a 'self-correction' update (Eqs. 14-15) that is supposed to realign previously edited parameters to the current edit. The oldest parameter is dequeued based on a threshold. Experiments on ZSRE, CounterFact, and RIPE with GPT-J (6B), LLaMA3 (8B), and Qwen2.5-14B report that QueueEDIT outperforms baselines at 1000 edits and remains competitive in single-turn editing, while preserving general capabilities on CSQA, MMLU, ANLI, and SQuAD2.

Significance. The paper addresses a real and important problem: sequential model editing in LLMs, where parameter updates can interfere with earlier edits and degrade general capabilities. The empirical backbone is extensive: multiple backbones, three editing benchmarks, edit counts of 1, 10, 100, 1000, an ablation study, a queue-length memory analysis, and a larger-backbone experiment with Qwen2.5-14B. The reported gains over DAFNet, the strongest baseline, are substantial at 1000 edits (e.g., 69.1 vs 65.1 average on ZSRE, 61.8 vs 55.3 on CounterFact, 61.2 vs 55.6 on RIPE), and the queue-based self-correction idea is a sensible approach to modeling dependencies among sequential edits. If the method is properly specified and reproducible, these results would be a useful contribution to the SME literature. However, the central self-correction update is not well-defined as written, and the similarity assumption underlying the top-K selection is only anecdotally validated, so the significance is currently contingent on resolving these specification gaps.

major comments (4)
  1. [§3.3, Eqs. (14)-(15)] The self-correction update is not a well-defined tensor operation as written. In Eq. (14), ΔW is defined as an element-wise sum of two vectors v_t^* and h_i^r, so it is a vector. In Eq. (15), this vector is concatenated with W_i (the located parameter matrix) via the symbol ∥, and then σ and b' are introduced without definition. A matrix cannot be concatenated with a vector without specifying the concatenation axis, the reshaping or broadcasting rule for the vector, and the shapes of σ and b'; moreover, the write-back of W_i' 'into the LLM' is not described. Since this step is the entire mechanism claimed to produce the performance gains, the paper must provide a precise, reproducible specification of Eqs. (14)-(15), including tensor shapes, or provide pseudocode and released code. As it stands, the reported results rest on an unimplementable step.
  2. [§3.3, Eq. (12)] The Euclidean distance between located parameter matrices is assumed to be a valid proxy for semantic relatedness between the corresponding knowledge triples, and this assumption is load-bearing for the top-K selection and the threshold-based dequeue. The paper only provides qualitative case studies in Appendix B.2 and Figure 5; no quantitative evidence is given that parameter-space distance tracks fact similarity. A concrete validation would be to compare the distance-based top-K selection against an oracle that knows the true co-occurrence structure of the edit sequence, or to correlate the distances in Eq. (12) with an external semantic similarity measure (e.g., relation/entity embedding cosine similarity) on held-out edit sequences. Without such evidence, the apparent gains could be an artifact of the specific benchmark distributions.
  3. [§3.2 and §3.4] The relationship between the closed-form ROME update (Eq. 7), the structural editing loss (Eq. 11), and the total training loss L_total in §3.4 is unspecified. The paper says parameters are 'derived via a closed-form solution' (Eq. 7) but also says the gradient of L_st is used to optimize W_proj^{l0}, and L_total combines L_ed and L_st. It is unclear whether the closed-form update is applied first, whether L_total is optimized from the closed-form initialization, or whether the two are alternative procedures. The order of operations for each edit (closed-form update, gradient-based loss optimization, queue enqueue/top-K/self-correction/dequeue) must be stated precisely for the method to be reproducible.
  4. [§4.2 and §A.3] The main results depend on several hyperparameters that are only partially analyzed. Appendix A.3 fixes K=50, eta_que=0.5, eta_deq=0.5, and the loss coefficients alpha_1 and alpha_2, but the only sensitivity study reported is queue length (Table 2). No experiments vary K or the two thresholds, and the loss coefficients are not discussed at all. Because the performance advantage of QueueEDIT could hinge on these choices, the paper should report sensitivity of the main claims to K and the thresholds, or provide a principled justification for the chosen values beyond 'limitations of machine resources.'
minor comments (5)
  1. [Footnote 1] The paper states that source code will be released upon acceptance but provides no code or pseudocode now; given the specification gaps in Eqs. (14)-(15), a detailed algorithm listing in the paper or an anonymous code repository would substantially aid reproducibility.
  2. [Table 2] The memory consumption comparison in Table 2 reports memory for the baselines but does not report QueueEDIT's own memory usage, even though the text states that QueueEDIT 'consumes slightly more memory'; please add the QueueEDIT memory numbers to the table.
  3. [§3.3, Step 3] The sentence 'In general, it suffices to replace the object of the t-th sample with the subject of the i-th sample' is confusing and appears to describe the semantic relation between edits rather than a concrete operation; please rewrite it to state unambiguously what is replaced and how it relates to Eqs. (14)-(15).
  4. [Appendix A.3] The training procedure states 'When the number of sequential editing models reaches the maximum value, we perform an additional 20,000 iterations before stopping'; it is unclear whether these additional iterations are on the same edit sequence or on new synthetic facts, and this detail affects reproducibility.
  5. [Figure 3] Figure 3 reports average general-capability accuracy without error bars or statistical significance tests; given the 'no degradation' claim in the abstract, please add variance information or a significance statement.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the method is trained on editing benchmarks and evaluated on held-out test splits, with no fitted quantity relabeled as a prediction.

full rationale

We walked the derivation chain in QueueEDIT. The structural mapping loss (Eq. 11) is a proposed training objective that adapts the TransE-style translation constraint to MLP hidden states (k_s^* W_proj + h_r - v*); it does not use a target quantity that was itself fitted from the same loss. The located parameters W_proj in Eq. 7 are inherited from ROME's closed-form editing solution (Meng et al. 2022), an external, independently published result, and the paper's own mechanism is the FIFO queue, top-K selection by Euclidean distance (Eq. 12), and the self-correction update (Eqs. 14-15). These components are evaluated on held-out editing sequences and against external benchmarks; the reliability/generality/locality metrics are computed on evaluation data, not on the training examples used to fit the queue updates. The paper cites the authors' prior DAFNet work for baselines and the DAFSet training set, and DAFNet is the strongest baseline, but the central claim does not reduce to that citation: QueueEDIT's gains are measured against DAFNet rather than derived from it. The main internal concern is not circularity but specification: Eq. 15 concatenates the matrix W_i with the vector ΔW without defining shapes, σ, or b', and footnote 1 states code will be released only upon acceptance. These are reproducibility gaps, not cases where a prediction reduces by construction to its input. No circular step can be exhibited with a quote-level equivalence, so the honest finding is no significant circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The weight queue is an algorithmic data structure, not a new physical or theoretical entity. No new particles, forces, dimensions, or conserved quantities are postulated. The free parameters listed above are all hand-chosen or selected by a small performance scan, and the central claim depends on them.

free parameters (5)
  • queue length ratio = 30% of 1000 edits
    The main experiments use 30% queue length; Table 2 reports 10%, 30%, and 50% and 30% gives the best average, so the value is selected by a performance scan.
  • top-K count K = 50
    K=50 is set 'due to limitations of machine resources' (Appendix A.3), not by validation; no sensitivity analysis is provided.
  • queue similarity threshold eta_que = 0.5
    Hand-set threshold in Appendix A.3 to decide whether a queued parameter is semantically correlated with the current edit.
  • dequeue threshold eta_deq = 0.5
    Hand-set threshold in Appendix A.3 to decide whether the queue head is far enough to evict.
  • loss coefficients alpha_1 and alpha_2 = not reported
    The total loss L_total = alpha_1 * L_ed + alpha_2 * L_st with alpha_1 + alpha_2 = 1; the actual values are not stated in the paper.
assumptions (4)
  • domain assumption The FFN layer l0 selected by causal tracing is the correct layer for editing (Meng et al. 2022, 2023; Zhang et al. 2024).
    QueueEDIT inherits ROME's locate-then-edit premise without re-deriving it for the new structural loss, invoked in Section 3.2 'Locating the MLP Layer'.
  • domain assumption Knowledge triple semantics follow the translation model s + r approximately equals o (Bordes et al. 2013), and this translation holds in the MLP activation space.
    Eq. 11 defines the structural loss as ||k_s* W_proj + h_r - v*||^2, assuming subject and relation representations combine linearly toward the object representation inside the FFN.
  • domain assumption Updating only W_proj of the located layer is sufficient for reliability, generality, and locality.
    The closed-form update in Eq. 7 and the structural loss only optimize W_proj; the possibility that other layers should be adjusted is not tested.
  • standard math The constant C = K * K^T from cached Wikipedia embeddings is a valid statistics matrix for these backbones.
    Inherited from ROME (Eq. 7); the paper does not recompute or validate this constant for GPT-J or LLaMA3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of QueueEDIT: Structural Self-Correction for Sequential Model Editing in LLMs." pith.science (2026). https://pith.science/paper/HDK6DPVI

@misc{pith2026250617864,
  author       = {Pith},
  title        = {Pith review of: QueueEDIT: Structural Self-Correction for Sequential Model Editing in LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HDK6DPVI}},
  note         = {Machine review of arXiv:2506.17864}
}
read the original abstract

Recently, large language models (LLMs) have demonstrated impressive results but still suffer from hallucinations. Model editing has been proposed to correct factual inaccuracies in LLMs. A challenging case is sequential model editing (SME), which aims to rectify errors continuously rather than treating them as a one-time task. During SME, the general capabilities of LLMs can be negatively affected due to the introduction of new parameters. In this paper, we propose a queue-based self-correction framework (QueueEDIT) that not only enhances SME performance by addressing long-sequence dependency but also mitigates the impact of parameter bias on the general capabilities of LLMs. Specifically, we first introduce a structural mapping editing loss to map the triplets to the knowledge-sensitive neurons within the Transformer layers of LLMs. We then store the located parameters for each piece of edited knowledge in a queue and dynamically align previously edited parameters. In each edit, we select queue parameters most relevant to the currently located parameters to determine whether previous knowledge needs realignment. Irrelevant parameters in the queue are frozen, and we update the parameters at the queue head to the LLM to ensure they do not harm general abilities. Experiments show that our framework significantly outperforms strong baselines across various SME settings and maintains competitiveness in single-turn editing. The resulting LLMs also preserve high capabilities in general NLP tasks throughout the SME process.

Figures

Figures reproduced from arXiv: 2506.17864 by the authors.

Figure 1
Figure 1. Comparison between SME methods. Modifying Parameters and Adding Extra Parameters ap￾proaches change parameters at specific positions in FFN layers without making corresponding adjustments for relevant edits. Retrieving Data methods focus solely on external data for editing knowledge triples. QueueEDIT not only dynamically updates parameters associated with previous edits but also ensures the general capabilities of … view at source ↗
Figure 2
Figure 2. The QueueEDIT framework. (a) Structural Mapping Editing maps each triple (s, r, o) to dedicated positions (i.e., k s ∗ , hr, and v∗) within the FFN layer. (b) Queue-based Self-Correction stores the located parameters into a Weight Queue following the FIFO principle for subsequent semantic alignment, aiming to preserve the general capabilities of LLMs. where Λ = (v∗−W l0 projk∗) (C−1k∗)T k∗ and C = K ·KT is a constan… view at source ↗
Figure 3
Figure 3. The “improvement/degeneration” of general capa [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Results using Qwen2.5-14B as the backbone. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative analysis of queue-based self-correction in two case studies. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 34 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    An, S.; Ma, Z.; Lin, Z.; Zheng, N.; Lou, J.; and Chen, W. 2023. Learning From Mistakes Makes LLM Better Reasoner. CoRR, abs/2310.20689

    arXiv 2023

  4. [4]

    R.; and Casas, N

    Basta, C.; Costa - juss \` a , M. R.; and Casas, N. 2021. Extensive study on the underlying gender bias in contextualized word embeddings. Neural Comput. Appl., 3371--3384

    work page 2021

  5. [5]

    Blinova, S.; Zhou, X.; Jaggi, M.; Eickhoff, C.; and Bahrainian, S. A. 2023. SIMSUM: Document-level Text Simplification via Simultaneous Summarization. In ACL, 9927--9944

    work page 2023

  6. [6]

    Bordes, A.; Usunier, N.; Garc \' a - Dur \' a n, A.; Weston, J.; and Yakhnenko, O. 2013. Translating Embeddings for Modeling Multi-relational Data. In NIPS, 2787--2795

    work page 2013

  7. [7]

    D.; Aziz, W.; and Titov, I

    Cao, N. D.; Aziz, W.; and Titov, I. 2021. Editing Factual Knowledge in Language Models. In Moens, M.; Huang, X.; Specia, L.; and Yih, S. W., eds., EMNLP, 6491--6506

    work page 2021

  8. [8]

    Chen, Q.; Zhang, T.; He, X.; Li, D.; Wang, C.; Huang, L.; and Xue', H. 2024. Lifelong Knowledge Editing for LLMs with Retrieval-Augmented Continuous Prompt Learning. In EMNLP, 13565--13580

    work page 2024

Show all 52 references
  1. [9]

    Z.; and Fu, J

    Chenmien Tan, G. Z.; and Fu, J. 2023. Massive Editing for Large Language Models via Meta Learning. CoRR, abs/2311.04661

    2023 arXiv

  2. [10]

    Cohen, R.; Biran, E.; Yoran, O.; Globerson, A.; and Geva, M. 2023. Evaluating the Ripple Effects of Knowledge Editing in Language Models. CoRR, abs/2307.12976

    2023 arXiv

  3. [11]

    Dai, D.; Dong, L.; Hao, Y.; Sui, Z.; Chang, B.; and Wei, F. 2022. Knowledge Neurons in Pretrained Transformers. In ACL, 8493--8502

    2022

  4. [12]

    Dong, Q.; Dai, D.; Song, Y.; Xu, J.; Sui, Z.; and Li, L. 2022. Calibrating Factual Knowledge in Pretrained Language Models. In Goldberg, Y.; Kozareva, Z.; and Zhang, Y., eds., EMNLP, 5937--5947

    2022

  5. [13]

    Dubey, A.; Jauhri, A.; Pandey, A.; Kadian, A.; Al - Dahle, A.; Letman, A.; Mathur, A.; Schelten, A.; Yang, A.; Fan, A.; Goyal, A.; Hartshorn, A.; Yang, A.; Mitra, A.; Sravankumar, A.; Korenev, A.; Hinsvark, A.; Rao, A.; Zhang, A.; Rodriguez, A.; Gregerson, A.; Spataru, A.; Roz...

    2024 arXiv

  6. [14]

    Fang, J.; Jiang, H.; Wang, K.; Ma, Y.; Shi, J.; Wang, X.; He, X.; and Chua, T. 2025. AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models. In ICLR

    2025

  7. [15]

    Gupta, A.; and Anumanchipalli, G. 2024. Rebuilding ROME : Resolving Model Collapse during Sequential Model Editing. CoRR, abs/2403.07175

    2024 arXiv

  8. [16]

    Gupta, A.; Rao, A.; and Anumanchipalli, G. 2024. Model Editing at Scale leads to Gradual and Catastrophic Forgetting. In ACL, 15202--15232

    2024

  9. [17]

    Han, X.; Li, R.; Tan, H.; Yuanlong, W.; Chai, Q.; and Pan, J. 2023. Improving Sequential Model Editing with Fact Retrieval. In EMNLP, 11209--11224

    2023

  10. [18]

    Hartvigsen, T.; Sankaranarayanan, S.; Palangi, H.; Kim, Y.; and Ghassemi, M. 2022. Aging with GRACE: Lifelong Model Editing with Discrete Key-Value Adaptors. CoRR, abs/2211.11031

    2022 arXiv

  11. [19]

    Hendrycks, D.; Burns, C.; Basart, S.; Zou, A.; Mazeika, M.; Song, D.; and Steinhardt, J. 2021. Measuring Massive Multitask Language Understanding. In ICLR

    2021

  12. [20]

    Hu, C.; Cao, P.; Chen, Y.; Liu, K.; and Zhao, J. 2024. WilKE: Wise-Layer Knowledge Editor for Lifelong Knowledge Editing. CoRR, abs/2402.10987

    2024 arXiv

  13. [21]

    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. 2022. LoRA: Low-Rank Adaptation of Large Language Models. In ICLR

    2022

  14. [22]

    Huang, Z.; Shen, Y.; Zhang, X.; Zhou, J.; Rong, W.; and Xiong, Z. 2023. Transformer-Patcher: One Mistake Worth One Neuron. In ICLR

    2023

  15. [23]

    H \" u botter, J.; Bongni, S.; Hakimi, I.; and Krause, A. 2025. Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs. In ICLR

    2025

  16. [24]

    Jiang, Y.; Wang, Y.; Wu, C.; Zhong, W.; Zeng, X.; Gao, J.; Li, L.; Jiang, X.; Shang, L.; Tang, R.; Liu, Q.; and Wang, W. 2024. Learning to Edit: Aligning LLMs with Knowledge Editing. CoRR, abs/2402.11905

    2024 arXiv

  17. [25]

    Kim, K.; Jang, J.; Choi, J.; Lee, Y.; Jin, K.; and Kim, Y. 2025. Plug-in and Fine-tuning: Bridging the Gap between Small Language Models and Large Language Models

    2025

  18. [26]

    Krishna, S.; Singh, B.; Roychowdhury, S.; Sridhara, G.; Mazumdar, S.; Sandelin, M.; Rentas, D.; Nalepa, M.; Sawicki, K.; and Gajda, J. 2025. Test code generation at Ericsson using Program Analysis Augmented Fine Tuned LLMs

    2025

  19. [27]

    Levy, O.; Seo, M.; Choi, E.; and Zettlemoyer, L. 2017. Zero-Shot Relation Extraction via Reading Comprehension. In CoNLL, 333--342

    2017

  20. [28]

    Lewis, M.; Liu, Y.; Goyal, N.; Ghazvininejad, M.; Mohamed, A.; Levy, O.; Stoyanov, V.; and Zettlemoyer, L. 2020. BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension. In ACL, 7871--7880

    2020

  21. [29]

    Li, Y.; Dong, B.; Guerin, F.; and Lin, C. 2023. Compressing Context to Enhance Inference Efficiency of Large Language Models. In EMNLP, 6342--6353

    2023

  22. [30]

    Lin, Y.; Liu, Z.; Sun, M.; Liu, Y.; and Zhu, X. 2015. Learning Entity and Relation Embeddings for Knowledge Graph Completion. In AAAI, 2181--2187

    2015

  23. [31]

    Liu, D.; Chen, M.; Lu, B.; Jiang, H.; Han, Z.; Zhang, Q.; Chen, Q.; Zhang, C.; Ding, B.; Zhang, K.; Chen, C.; Yang, F.; Yang, Y.; and Qiu, L. 2024. RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval. CoRR, abs/2409.10516

    2024 arXiv

  24. [32]

    Madaan, A.; Tandon, N.; Clark, P.; and Yang, Y. 2022. Memory-assisted prompt editing to improve GPT-3 after deployment. In EMNLP, 2833--2861

    2022

  25. [33]

    Meng, K.; Bau, D.; Andonian, A.; and Belinkov, Y. 2022. Locating and Editing Factual Associations in GPT . In NeurIPS

    2022

  26. [34]

    S.; Andonian, A

    Meng, K.; Sharma, A. S.; Andonian, A. J.; Belinkov, Y.; and Bau, D. 2023. Mass-Editing Memory in a Transformer. In ICLR

    2023

  27. [35]

    Mishra, A.; Asai, A.; Balachandran, V.; Wang, Y.; Neubig, G.; Tsvetkov, Y.; and Hajishirzi, H. 2024. Fine-grained Hallucination Detection and Editing for Language Models. CoRR, abs/2401.06855

    2024 arXiv

  28. [36]

    Mitchell, E.; Lin, C.; Bosselut, A.; Finn, C.; and Manning, C. D. 2022 a . Fast Model Editing at Scale. In ICLR

    2022

  29. [37]

    D.; and Finn, C

    Mitchell, E.; Lin, C.; Bosselut, A.; Manning, C. D.; and Finn, C. 2022 b . Memory-Based Model Editing at Scale. In ICML, 15817--15831

    2022

  30. [38]

    Nie, Y.; Williams, A.; Dinan, E.; Bansal, M.; Weston, J.; and Kiela, D. 2020. Adversarial NLI: A New Benchmark for Natural Language Understanding. In Jurafsky, D.; Chai, J.; Schluter, N.; and Tetreault, J. R., eds., ACL, 4885--4901

    2020

  31. [39]

    Qwen Team . 2024. Qwen2.5: A Party of Foundation Models

    2024

  32. [40]

    Rajpurkar, P.; Zhang, J.; Lopyrev, K.; and Liang, P. 2016. SQuAD: 100, 000+ Questions for Machine Comprehension of Text. In EMNLP, 2383--2392

    2016

  33. [41]

    Reimers, N.; and Gurevych, I. 2019. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. In Inui, K.; Jiang, J.; Ng, V.; and Wan, X., eds., EMNLP, 3980--3990

    2019

  34. [42]

    M.; Sankaranarayanan, K.; and Chandar, S

    Saha, A.; Pahuja, V.; Khapra, M. M.; Sankaranarayanan, K.; and Chandar, S. 2018. Complex Sequential Question Answering: Towards Learning to Converse Over Linked Question Answer Pairs with a Knowledge Graph. In AAAI, 705--713

    2018

  35. [43]

    H.; Sch \" a rli, N.; and Zhou, D

    Shi, F.; Chen, X.; Misra, K.; Scales, N.; Dohan, D.; Chi, E. H.; Sch \" a rli, N.; and Zhou, D. 2023. Large Language Models Can Be Easily Distracted by Irrelevant Context. In ICML, 31210--31227

    2023

  36. [44]

    Tam, D.; Mascarenhas, A.; Zhang, S.; Kwan, S.; Bansal, M.; and Raffel, C. 2023. Evaluating the Factual Consistency of Large Language Models Through News Summarization. In ACL, 5220--5255

    2023

  37. [45]

    Wang, P.; Zhang, N.; Xie, X.; Yao, Y.; Tian, B.; Wang, M.; Xi, Z.; Cheng, S.; Liu, K.; Zheng, G.; and Chen, H. 2023. EasyEdit: An Easy-to-use Knowledge Editing Framework for Large Language Models. CoRR, abs/2308.07269

    2023 arXiv

  38. [46]

    Wang, Z.; Zhang, J.; Feng, J.; and Chen, Z. 2014. Knowledge Graph Embedding by Translating on Hyperplanes. In AAAI, 1112--1119

    2014

  39. [47]

    Yao, Y.; Wang, P.; Tian, B.; Cheng, S.; Li, Z.; Deng, S.; Chen, H.; and Zhang, N. 2023. Editing Large Language Models: Problems, Methods, and Opportunities. CoRR, abs/2305.13172

    2023 arXiv

  40. [48]

    You, H.; Guo, Y.; Fu, Y.; Zhou, W.; Shi, H.; Zhang, X.; Kundu, S.; Yazdanbakhsh, A.; Yingyan; and Lin. 2024. ShiftAddLLM: Accelerating Pretrained LLMs via Post-Training Multiplication-Less Reparameterization. arXiv:2406.05981

    2024 arXiv

  41. [49]

    Yu, L.; Chen, Q.; Zhou, J.; and He, L. 2024. MELO: Enhancing Model Editing with Neuron-Indexed Dynamic LoRA. In AAAI, 19449--19457

    2024

  42. [50]

    Zhang, T.; Chen, Q.; Li, D.; Wang, C.; He, X.; Huang, L.; Xue', H.; and Huang, J. 2024. DAFNet: Dynamic Auxiliary Fusion for Sequential Model Editing in Large Language Models. In ACL, 1588--1602

    2024

  43. [51]

    Zheng, C.; Li, L.; Dong, Q.; Fan, Y.; Wu, Z.; Xu, J.; and Chang, B. 2023. Can We Edit Factual Knowledge by In-Context Learning? CoRR, abs/2305.12740

    2023 arXiv

  44. [52]

    Zheng, H.; Zhang, X.; Chi, Z.; Huang, H.; Tan, Y.; Lan, T.; Wei, W.; and Mao, X. 2022. Cross-Lingual Phrase Retrieval. In ACL, 4193--4204

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.