Pith. sign in

REVIEW 4 major objections 5 minor 41 references

Exploiting Contextual Knowledge in LLMs through V-usable Information based Layer Enhancement

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Amplifying one mid-layer hidden state makes LLMs answer from context, not memory.

desk verdict A simple, training-free layer intervention that consistently improves context-faithful QA, but the stated mechanism is shakier than the results. read the letter →

arxiv 2504.15630 v1 pith:MLVVXKTY submitted 2025-04-22 cs.CL

Xiaowei Yuan , Zhao Yang , Ziyang Huang , Yequan Wang , Siqi Fan , Yiming Ju , Jun Zhao , Kang Liu This is my paper · ORCID
classification cs.CL
keywords context-faithfulnessV-usableinformationlayerenhancementknowledgeconflictshiddenstateinterventionlogitlensquestionansweringlargelanguagemodels
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

Large language models often answer from memory instead of the supplied context, and existing fixes work at the decoding level without touching how context moves through the model. This paper argues that the context-correct answer is actually present in an intermediate layer's hidden state, then gets diluted in deeper layers, and introduces CaLE, an inference-time intervention that finds that layer and amplifies it. Using V-usable information to pick the layer, CaLE scales the hidden state (or adds a residual connection) so the final representation is enriched with contextual knowledge. On QA benchmarks with unknown or conflicting contexts, CaLE reports exact-match improvements over original decoding, early exit, and IRCAN across multiple Llama, Mistral, and Gemma models, and it stacks with contrastive decoding strategies.

What carries the argument

The machinery is V-usable information estimated by a logit lens: $I_V(h_l\to Y)=H_V(Y)-H_V(Y|h_l)$, where the conditional V-entropy is computed from a softmax over logit-lens logits at layer $l$. This identifies an optimal layer before contextual information plateaus or decays. The intervention uses the residual-stream decomposition of final logits as $v+u(v)$, where $v=h_l W_U$ and $u(v)$ collects contributions from later layers; amplifying to $\alpha h_l$ turns the final logits into $\alpha v+u(\alpha v)$. Proposition 3.1 states that if $k=\arg\max_j v_j$, then $\lim_{\alpha\to\infty} H_{V,f}(\alpha)\approx 0$, so the final distribution collapses onto the context-correct token. The residual-connection variant is shown to behave like an effective $(\alpha+1)$ scaling.

What would settle it

Take a CounterFact example whose logit lens at the layer CaLE selects ranks the parametric wrong answer above the context-correct answer, then apply CaLE-A at $\alpha = 2,4,8,16$; if the final softmax probability of the wrong answer rises with $\alpha$, the premise that the correct token is the layer's argmax is violated and the claimed mechanism is refuted for that case.

Watch

Extended reading notes

Core claim

The core claim is that context-faithfulness is impaired not by a lack of contextual information but by its attenuation in deeper layers: the logit-lens distribution at some intermediate layer already ranks the context-correct answer near the top, and later layers dilute it. CaLE therefore locates that layer, using V-usable information in a supervised setting and a KL-divergence proxy in an unsupervised setting, and amplifies its hidden states (CaLE-A) or adds a residual connection from it (CaLE-R). The paper proves that if the correct token is the argmax of the layer-l logit vector, then as the amplification factor grows, the conditional V-entropy at the final layer tends to zero, so the final softmax concentrates on the context answer. Experiments on CounterFact, NQ, NQ-Swap, SQuAD, and StrategyQA show exact-match, F1, or accuracy gains over original decoding, early exit, and IRCAN across Llama, Mistral, and Gemma models, with the largest gains when the context conflicts with parametric knowledge.

Load-bearing premise

CaLE assumes the layer it amplifies already ranks the context-correct answer at or near the top of its logit-lens distribution and that the scaled hidden state reaches the final logits linearly; when the parametric wrong answer is on top, amplification strengthens the wrong prediction.

Editorial extensions

If this is right

  • CaLE-A and CaLE-R both improve exact match over original decoding, early exit, and IRCAN on CounterFact, with the largest gains on the subset where context contradicts parametric knowledge.
  • On NQ-Swap, a dataset built entirely from conflicting contexts, CaLE raises exact match substantially across Llama, Mistral, and Gemma models, showing the intervention specifically helps when context should override memory.
  • Unsupervised layer selection via KL divergence matches or approaches supervised selection and still beats all baselines, so CaLE does not require labeled validation data.
  • CaLE composes with contrastive decoding methods such as CAD, CD, and COIECD, yielding cumulative gains over either intervention alone.
  • The best intervention layer is consistently in the middle-to-deep range, and amplifying hidden states helps while amplifying attention or MLP outputs does not.

Reading between the lines

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

  • An extension the paper leaves implicit is a per-instance layer selector: compute the KL-based proxy on each input and amplify only when the context-correct token is already near the top of that layer's logit-lens distribution, making the intervention adaptive rather than dataset-wide.
  • Because CaLE is orthogonal to decoding strategies, a natural next test is combining it with neuron-reweighting methods such as IRCAN rather than only with contrastive decoding; the paper tests CaLE with CAD, CD, and COIECD but not with IRCAN plus CaLE.
  • If the mechanism holds, it offers a diagnostic: an LLM that ignores context may have the correct answer localized at a specific layer and then diluted, so probing that layer could predict which inputs will be answered from memory rather than from the supplied evidence.
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. The paper proposes CaLE, a training-free inference-time intervention for improving context-faithful QA. CaLE identifies a 'context-aware layer' using either a supervised validation-accuracy criterion (Eq. 11) or an unsupervised KL-divergence criterion (Eq. 18), then enhances that layer's hidden states by amplifying them (CaLE-A) or adding a residual path (CaLE-R) before continuing the forward pass. The method is evaluated on CounterFact, NQ, NQ-Swap, SQuAD, and StrategyQA across Llama, Mistral, and Gemma models, reporting consistent EM/F1 improvements over original decoding, early exit, and IRCAN, with further gains when combined with contrastive decoding methods. The theoretical support (Proposition 3.1) claims that the final-layer conditional V-entropy tends to zero when the chosen layer's logit-lens distribution has the correct answer token at its maximum.

Significance. The empirical finding is potentially useful: if the reported gains hold, CaLE offers a simple, plug-in intervention that improves context faithfulness without fine-tuning, and its orthogonality to contrastive decoding is a practical advantage. Strengths include the breadth of the evaluation (five datasets, five model families), the ablation over alpha and intervention location, and the layer-consistency analysis for supervised selection. However, the paper's stated mechanism is not established: the formal guarantee hinges on a condition the authors admit is unverifiable, and the correlation supporting the unsupervised metric is not quantified. The independent EM/F1 evidence is the main support, and it could be made stronger with error bars and a clearly described hyperparameter selection protocol.

major comments (4)
  1. [Section 3.1.1 (Prop. 3.1; App. E)] Proposition 3.1 requires k = arg max_j v_j at the chosen layer, a condition that the paper explicitly states cannot be guaranteed in Section 3.1.1, and neither the supervised layer selector in Eq. (11) nor the unsupervised KL_c selector in Eq. (18) checks the rank of the context-correct answer token at the selected layer. For any example where the correct token is not top-1 at layer l, amplifying by alpha > 1 sharpens the wrong token, so the stated mechanism predicts degradation rather than improvement; the aggregate gains in Tables 1 and 3 therefore remain unexplained at the per-example level. I ask the authors to report the frequency with which the argmax condition holds at the selected layers, or to replace the global theoretical claim with a conditional statement and empirical diagnostics.
  2. [Appendix C and Appendix E] The proof of Proposition 3.1 is incomplete as written. Appendix C ignores the final LayerNorm, and Appendix E asserts that the terms u_j(alpha v_j) - u_k(alpha v_k) are negligible without providing a bound; footnote 5 states 'Since alpha approaches zero' when the limit taken is alpha -> infinity. Because the subsequent-layer contributions u(alpha v) may themselves grow with alpha, the claimed exponential decay of the j != k terms is not established. The authors should provide a rigorous treatment of the u terms, for example by bounding |u_j(alpha v) - u_k(alpha v)| relative to alpha(v_j - v_k), or state the additional assumptions needed for the proof to go through.
  3. [Section 4.1, footnote; Figure 6] The hyperparameters alpha1 and alpha2 are listed as fixed values (5/3 for CounterFact and 3/1 for other datasets) without any statement that they were selected on a held-out set. The ablation in Figure 6 shows that performance on CounterFact varies substantially with alpha, and the chosen alpha1 = 5 sits at the upper edge of the best-performing range, suggesting that the test set was used to select alpha. If so, the results in Tables 1 and 3 are optimistically biased. Please report the selection procedure explicitly or provide a per-dataset sensitivity analysis demonstrating that the improvements are robust across a range of alpha values.
  4. [Tables 1-3] The main empirical claims are made without error bars, confidence intervals, or significance tests. Some of the improvements over the strongest baseline are modest (e.g., Table 3, SQuAD for Llama3.1-8B: supervised CaLE-A EM 67.38 vs. IRCAN 64.58), and the reader cannot tell whether these differences are consistent across examples. Reporting bootstrap intervals or paired tests over the test items would materially strengthen the paper's central claim.
minor comments (5)
  1. [Throughout] There are numerous typos: 'Unkown' (Appendix F.1), 'almostly' (Section 4.3), 'negative effective' (Section 1), 'lay' instead of 'layer' (Section 3.2.1), and 'faithfulnes' (Section 3).
  2. [Figure 2] The caption says '-HV' but the axis label is '-H(Y|h)'; please unify the notation.
  3. [Eq. (7)] The symbol v is reused in Eq. (7) for the logits at layer l after being defined in Eq. (4) for a generic layer; clarify the notation to avoid confusion.
  4. [Section 3.2.2] The unsupervised KL_c criterion in Eq. (18) is evaluated on the test set inputs; this transductive use of test inputs should be acknowledged, and the paper should state whether the layer would be selected from a separate unlabeled set in a deployment scenario.
  5. [Appendix E] The footnote in the proof says 'Since alpha approaches zero' but the limit is alpha -> infinity; this appears to be a typo and should be corrected.

Circularity Check

2 steps flagged · score 2.0 of 10

Minor circular validation in the internal-metric analysis; held-out EM/F1 keeps the central claim independent.

  1. other [Section 3.2.2 (Eq. 18) and Section 4.2.3 (Figure 5b)]
    "l∗ = arg max_l E_{P(q,c)}[KLc(l)] (18) ... The peak observed at layer 25 aligns with the layer selected by the CaLE method, offering robust validation for both approaches."

    For the unsupervised variant, Eq. 18 defines the selected layer as the argmax of the average KLc over layers. The KLc curve therefore peaks at the selected layer by construction. Citing that peak as 'robust validation for both approaches' is, for the unsupervised method, a restatement of the selection rule rather than an independent check. The supervised variant does provide independent support because that layer is chosen by validation accuracy, but the sentence credits both methods with the same evidence.

  2. self definitional [Section 3.1.1 (Proposition 3.1) and Section 4.2.3 (Figure 5a)]
    "If k = arg max_j v_j, then lim_{α→∞} H_Vf(α)≈ 0 ... Figure 5a provides theoretical validation for the effectiveness of our CaLE-A method."

    The V-usable information metric is defined via the logit-lens softmax probability of the answer token Y (Eq. 4). Proposition 3.1 is a softmax-concentration theorem: alpha-scaling a layer whose logit vector is topped by Y drives the conditional V-entropy to zero. The observed 'increase in V-usable information' in Figure 5a is therefore the theorem's conclusion, not an independent measurement of contextual-information growth whenever the premise k=argmax holds. The paper concedes it 'cannot guarantee' this premise, and held-out EM/F1 evaluations do provide independent evidence, so this is a partial, non-central circularity.

full rationale

The central claim that CaLE improves context-faithful QA is supported by held-out Exact Match and F1 evaluations on CounterFact, NQ, NQ-Swap, SQuAD, and StrategyQA against original decoding, early exit, and IRCAN baselines. Those results are not determined by the V-usable information definition. The supervised layer selection (Eq. 11) is standard validation-set tuning, and the unsupervised KLc selection (Eq. 18) is an approximation justified by correlation rather than by identity. The main circularity concerns are internal: (1) the unsupervised method's own KLc peak is used as 'validation' although the peak is its selection criterion by construction; and (2) the theoretical guarantee (Proposition 3.1) is a softmax-concentration statement about the very metric used to measure 'contextual information,' so Figure 5a's improved V-entropy is a theorem restatement conditional on an unverified per-example premise. The paper explicitly acknowledges this premise cannot be guaranteed. Self-citations (e.g., Yuan et al. 2024, Fan et al. 2024) are baselines or related work and are not load-bearing. No uniqueness theorem or ansatz is smuggled in via self-citation. The held-out benchmark comparisons keep the central empirical claim independent, so the overall circularity is minor.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical or model-level entities. The main hidden costs are analytical assumptions (logit-lens faithfulness, residual-stream linearity, the KL surrogate, the argmax condition) and per-dataset tuning parameters alpha1 and alpha2. These are counted in the free-parameter and axiom lists.

free parameters (3)
  • Amplification factor alpha1 (CaLE-A) = 5 on CounterFact; 3 on NQ, SQuAD, StrategyQA; ablation sweeps 0.5 to 16
    Set per dataset with no formal tuning protocol; the Figure 6 ablation is used to justify the CounterFact value.
  • Residual span alpha2 (CaLE-R) = 3 on CounterFact; 1 on other datasets
    Chosen per dataset; no search procedure described, making it a hand-set free parameter.
  • Validation set size for supervised layer selection = 500 samples
    Hand-selected; Figure 4a argues stability across 20 trials, but 500 is not derived from the problem.
assumptions (5)
  • domain assumption Logit lens with softmax at each layer gives a faithful estimate of V-usable information and token probabilities (Eq. 4).
    Used in Eq. 1-4 to define IV and to select layers; not a formal guarantee for arbitrary hidden states.
  • domain assumption Final logits decompose linearly as h_l W_U plus later-layer contributions; the final LayerNorm is ignored (Appendix C, Eq. 20).
    Needed for Eq. 7 and Eq. 22; common in residual-stream analysis, but an idealization that the proof depends on.
  • domain assumption KL_c(l) approximates IV(h_l; Y), despite the bound in Eq. 14 being one-directional.
    The paper justifies this with empirical correlation in Figure 3, not with a mathematical guarantee.
  • ad hoc to paper At the chosen layer, the correct answer token is the argmax of the logit-lens distribution (condition in Proposition 3.1).
    The proof's limit result requires k = arg max_j v_j; the method does not verify this and falls back to a fixed alpha > 1 heuristic.
  • standard math Softmax sharpening identities used in the proof.
    The concentration of the softmax as alpha grows is a standard property, not a contribution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploiting Contextual Knowledge in LLMs through V-usable Information based Layer Enhancement." pith.science (2026). https://pith.science/paper/MLVVXKTY

@misc{pith2026250415630,
  author       = {Pith},
  title        = {Pith review of: Exploiting Contextual Knowledge in LLMs through V-usable Information based Layer Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MLVVXKTY}},
  note         = {Machine review of arXiv:2504.15630}
}
read the original abstract

Large Language Models (LLMs) have demonstrated remarkable capabilities in various tasks, yet they often struggle with context-faithfulness generations that properly reflect contextual knowledge. While existing approaches focus on enhancing the decoding strategies, they ignore the fundamental mechanism of how contextual information is processed within LLMs' internal states. As a result, LLMs remain limited in their ability to fully leverage contextual knowledge. In this paper, we propose Context-aware Layer Enhancement (CaLE), a novel intervention method that enhances the utilization of contextual knowledge within LLMs' internal representations. By employing V-usable information analysis, CaLE strategically amplifies the growth of contextual information at an optimal layer, thereby enriching representations in the final layer. Our experiments demonstrate that CaLE effectively improves context-faithful generation in Question-Answering tasks, particularly in scenarios involving unknown or conflicting contextual knowledge.

Figures

Figures reproduced from arXiv: 2504.15630 by the authors.

Figure 1
Figure 1. An illustration of CaLE Method. parametric knowledge (Xie et al., 2024). As illus￾trated in the upper part of [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Visualization of Information Flow. The ver [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Variation of the KL divergences across lay [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Validation set size impact on supervised layer [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visualization of Analysis on the CounterFact [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The effect of amplification across different [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 20 canonical work pages

  1. [1]

    Mitchell

    Amos Azaria and Tom M. Mitchell. 2023. The internal state of an LLM knows when its lying. CoRR, abs/2304.13734

    arXiv 2023

  2. [2]

    Baolong Bi, Shaohan Huang, Yiwei Wang, Tianchi Yang, Zihan Zhang, Haizhen Huang, Lingrui Mei, Junfeng Fang, Zehao Li, Furu Wei, Weiwei Deng, Feng Sun, Qi Zhang, and Shenghua Liu. 2024. https://arxiv.org/abs/2412.15280 Context-dpo: Aligning language models for context-faithfulness . Preprint, arXiv:2412.15280

    arXiv 2024

  3. [3]

    Chao Chen, Kai Liu, Ze Chen, Yi Gu, Yue Wu, Mingyuan Tao, Zhihang Fu, and Jieping Ye. 2024. INSIDE: llms' internal states retain the power of hallucination detection. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024

    work page 2024

  4. [4]

    Hovy, Hinrich Sch \" u tze, and Yoav Goldberg

    Yanai Elazar, Nora Kassner, Shauli Ravfogel, Abhilasha Ravichander, Eduard H. Hovy, Hinrich Sch \" u tze, and Yoav Goldberg. 2021. Measuring and improving consistency in pretrained language models. Trans. Assoc. Comput. Linguistics, 9:1012--1031

    work page 2021

  5. [5]

    Nelson Elhage, Neel Nanda, Catherine Olsson, Tom Henighan, Nicholas Joseph, Ben Mann, Amanda Askell, Yuntao Bai, Anna Chen, Tom Conerly, et al. 2021. A mathematical framework for transformer circuits. Transformer Circuits Thread, 1(1):12

    2021

  6. [6]

    Kawin Ethayarajh, Yejin Choi, and Swabha Swayamdipta. 2022. Understanding dataset difficulty with V-usable information. In International Conference on Machine Learning, ICML 2022, 17-23 July 2022, Baltimore, Maryland, USA , volume 162 of Proceedings of Machine Learning Research, pages 5988--6008. PMLR

    work page 2022

  7. [7]

    Siqi Fan, Xin Jiang, Xiang Li, Xuying Meng, Peng Han, Shuo Shang, Aixin Sun, Yequan Wang, and Zhongyuan Wang. 2024. Not all layers of llms are necessary during inference. CoRR, abs/2403.02181

    arXiv 2024

  8. [8]

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. 2024. https://arxiv.org/abs/2312.10997 Retrieval-augmented generation for large language models: A survey . Preprint, arXiv:2312.10997

    arXiv 2024

Show all 41 references
  1. [9]

    Mor Geva, Daniel Khashabi, Elad Segal, Tushar Khot, Dan Roth, and Jonathan Berant. 2021 a . Did aristotle use a laptop? A question answering benchmark with implicit reasoning strategies. Trans. Assoc. Comput. Linguistics, 9:346--361

    2021

  2. [10]

    Mor Geva, Roei Schuster, Jonathan Berant, and Omer Levy. 2021 b . Transformer feed-forward layers are key-value memories. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, EMNLP 2021, Virtual Event / Punta Cana, Dominican Republic, 7-11...

    2021

  3. [11]

    Michael Hanna, Ollie Liu, and Alexandre Variengien. 2023. How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Syst...

    2023

  4. [12]

    Ziwei Ji, Nayeon Lee, Rita Frieske, Tiezheng Yu, Dan Su, Yan Xu, Etsuko Ishii, Yejin Bang, Andrea Madotto, and Pascale Fung. 2023. Survey of hallucination in natural language generation. ACM Comput. Surv. , 55(12):248:1--248:38

    2023

  5. [13]

    Zhuoran Jin, Pengfei Cao, Hongbang Yuan, Yubo Chen, Jiexin Xu, Huaijun Li, Xiaojian Jiang, Kang Liu, and Jun Zhao. 2024. Cutting off the head ends the conflict: A mechanism for interpreting and mitigating knowledge conflicts in language models. In Findings of the Association f...

    2024

  6. [14]

    Tianjie Ju, Weiwei Sun, Wei Du, Xinwei Yuan, Zhaochun Ren, and Gongshen Liu. 2024. How large language models encode context knowledge? A layer-wise probing study. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Eva...

    2024

  7. [15]

    Kullback and R

    S. Kullback and R. A. Leibler. 1951. On information and sufficiency. The Annals of Mathematical Statistics, 22(1):79--86

    1951

  8. [16]

    Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, Kristina Toutanova, Llion Jones, Matthew Kelcey, Ming - Wei Chang, Andrew M

    Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur P. Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, Kristina Toutanova, Llion Jones, Matthew Kelcey, Ming - Wei Chang, Andrew M. Dai, Jakob Uszkoreit, Quoc Le, and...

    2019

  9. [17]

    Xiang Lisa Li, Ari Holtzman, Daniel Fried, Percy Liang, Jason Eisner, Tatsunori Hashimoto, Luke Zettlemoyer, and Mike Lewis. 2023. Contrastive decoding: Open-ended text generation as optimization. In Proceedings of the 61st Annual Meeting of the Association for Computational L...

    2023

  10. [18]

    Shayne Longpre, Kartik Perisetla, Anthony Chen, Nikhil Ramesh, Chris DuBois, and Sameer Singh. 2021. Entity-based knowledge conflicts in question answering. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, EMNLP 2021, Virtual Event / P...

    2021

  11. [19]

    Xin Men, Mingyu Xu, Qingyu Zhang, Bingning Wang, Hongyu Lin, Yaojie Lu, Xianpei Han, and Weipeng Chen. 2024. Shortgpt: Layers in large language models are more redundant than you expect. CoRR, abs/2403.03853

    2024 arXiv

  12. [20]

    Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov. 2022 a . Locating and editing factual associations in GPT . In Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processing Systems 2022, NeurIPS 2022, New Orleans, LA, USA,...

    2022

  13. [21]

    Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov. 2022 b . Locating and editing factual associations in GPT . In Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processing Systems 2022, NeurIPS 2022, New Orleans, LA, USA,...

    2022

  14. [22]

    nostalgebraist. 2020. Interpreting GPT : the logit lens. AI Alignment Forum

    2020

  15. [23]

    Catherine Olsson, Nelson Elhage, Neel Nanda, Nicholas Joseph, Nova DasSarma, Tom Henighan, Ben Mann, Amanda Askell, Yuntao Bai, Anna Chen, et al. 2022. In-context learning and induction heads. arXiv preprint arXiv:2209.11895

    2022 arXiv

  16. [24]

    Zexuan Qiu, Zijing Ou, Bin Wu, Jingjing Li, Aiwei Liu, and Irwin King. 2024. Entropy-based decoding for retrieval-augmented large language models. CoRR, abs/2406.17519

    2024 arXiv

  17. [25]

    Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. Squad: 100, 000+ questions for machine comprehension of text. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, EMNLP 2016, Austin, Texas, USA, November 1-4, 2016 ...

    2016

  18. [26]

    Ori Ram, Yoav Levine, Itay Dalmedigos, Dor Muhlgay, Amnon Shashua, Kevin Leyton-Brown, and Yoav Shoham. 2023. https://doi.org/10.1162/tacl_a_00605 In-context retrieval-augmented language models . Transactions of the Association for Computational Linguistics, 11:1316--1331

    2023 doi

  19. [27]

    Ruiyang Ren, Yuhao Wang, Yingqi Qu, Wayne Xin Zhao, Jing Liu, Hua Wu, Ji - Rong Wen, and Haifeng Wang. 2025. Investigating the factual knowledge boundary of large language models with retrieval augmentation. In Proceedings of the 31st International Conference on Computational ...

    2025

  20. [28]

    Dan Shi, Renren Jin, Tianhao Shen, Weilong Dong, Xinwei Wu, and Deyi Xiong. 2024 a . IRCAN: mitigating knowledge conflicts in LLM generation via identifying and reweighting context-aware neurons. CoRR, abs/2406.18406

    2024 arXiv

  21. [29]

    Weijia Shi, Xiaochuang Han, Mike Lewis, Yulia Tsvetkov, Luke Zettlemoyer, and Wen - tau Yih. 2024 b . Trusting your evidence: Hallucinate less with context-aware decoding. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational ...

    2024

  22. [30]

    Oscar Skean, Md Rifat Arefin, Yann LeCun, and Ravid Shwartz-Ziv. 2024. Does representation matter? exploring intermediate layers in large language models. In NeurIPs Workshop on Machine Learning and Compression

    2024

  23. [31]

    Zhongxiang Sun, Xiaoxue Zang, Kai Zheng, Yang Song, Jun Xu, Xiao Zhang, Weijie Yu, and Han Li. 2024. Redeep: Detecting hallucination in retrieval-augmented generation via mechanistic interpretability. CoRR, abs/2410.11414

    2024

  24. [32]

    Jian Xie, Kai Zhang, Jiangjie Chen, Renze Lou, and Yu Su. 2024. Adaptive chameleon or stubborn sloth: Revealing the behavior of large language models in knowledge conflicts. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-...

    2024

  25. [33]

    Ji Xin, Raphael Tang, Yaoliang Yu, and Jimmy Lin. 2021. Berxit: Early exiting for BERT with better fine-tuning and extension to regression. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, EACL 2021, O...

    2021

  26. [34]

    Yilun Xu, Shengjia Zhao, Jiaming Song, Russell Stewart, and Stefano Ermon. 2020. A theory of usable information under computational constraints. In 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020 . OpenReview.net

    2020

  27. [35]

    Qinan Yu, Jack Merullo, and Ellie Pavlick. 2023. Characterizing mechanisms for factual recall in language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023 , pages 9924--9959. Associati...

    2023

  28. [36]

    Xiaowei Yuan, Zhao Yang, Yequan Wang, Shengping Liu, Jun Zhao, and Kang Liu. 2024. Discerning and resolving knowledge conflicts through adaptive decoding with contextual information-entropy constraint. In Findings of the Association for Computational Linguistics, ACL 2024, Ban...

    2024

  29. [37]

    Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, Yifan Du, Chen Yang, Yushuo Chen, Zhipeng Chen, Jinhao Jiang, Ruiyang Ren, Yifan Li, Xinyu Tang, Zikang Liu, Peiyu Liu, Jian-Yun Nie, and Ji-Rong W...

    2024 arXiv

  30. [38]

    Hanzhang Zhou, Zijian Feng, Zixiao Zhu, Junlang Qian, and Kezhi Mao. 2024. Unibias: Unveiling and mitigating LLM bias through internal attention and FFN manipulation. In Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Sy...

    2024

  31. [39]

    Wenxuan Zhou, Sheng Zhang, Hoifung Poon, and Muhao Chen. 2023. Context-faithful prompting for large language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023 , pages 14544--14556. Association for Computational Linguistics

    2023

  32. [40]

    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...

  33. [41]

    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 16, 2026 · model on record in the stance chip above.