Pith. sign in

REVIEW 4 major objections 5 minor 49 references

Enhancing Uncertainty Modeling with Semantic Graph for Hallucination Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper claims that propagating uncertainty over a semantic graph of entities and sentences detects hallucinations better than scoring tokens independently, with a 19.78-point gain in passage-level detection on the Chinese dataset.

desk verdict Semantic graph idea is fresh but the evaluation is compromised by test-set tuning and manual graph curation; reject as submitted. read the letter →

arxiv 2501.02020 v3 pith:EFZKSEAQ submitted 2025-01-02 cs.CL cs.AI

Kedi Chen , Qin Chen , Jie Zhou , Xinqi Tao , Bowen Ding , Jingwen Xie , Mingchen Xie , Peilong Li
show 2 more authors
Feng Zheng Liang He
This is my paper · ORCID
classification cs.CLcs.AI
keywords hallucinationdetectionuncertaintyestimationsemanticgraphAbstractMeaningRepresentationpassage-levelnaturallanguageinferencelargemodelsfactuality
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

The paper tries to show that uncertainty-based hallucination detection fails when it treats each token as independent, and that a semantic graph of entity and sentence relations can fix that. The authors construct a passage-level graph by parsing each sentence into an Abstract Meaning Representation, linking entities across sentences, and then use the graph twice: to propagate uncertainty between related entities and to calibrate each sentence's score against contradicting neighbors. They report consistent gains over state-of-the-art baselines at both levels, with a 19.78-point improvement in passage-level Spearman correlation on a Chinese note-summarization dataset. If the claim holds, hallucination detection becomes more accurate without needing external knowledge bases or repeated sampling, which is what makes uncertainty-based detection attractive in the first place.

What carries the argument

The load-bearing object is the passage-level semantic graph, a graph built from AMR parses of each sentence, with Abstract Meaning Representation encoding who did what to whom and sentence graphs stitched together by coreference resolution and entity linking. Two mechanisms use it: relation-based uncertainty propagation, which routes uncertainty from subject to object entities along the dependency path, weighted by attention and penalized by relation intensity, and graph-based uncertainty calibration, which scales each sentence's uncertainty by the NLI contradiction probability with each neighboring sentence. The graph's role is to decide which tokens and sentences are actually related, so that uncertainty flows along true semantic connections rather than the work of scoring everything independently.

What would settle it

Run the published pipeline on a third dataset using only automatic graph construction, with no manual triple review; if the passage-level gain over FOCUS collapses to noise, the reported advantage depends on human cleanup rather than on the semantic-graph mechanism. A second check: build passages in which two sentences contradict each other but share no linked entity; if the graph-based calibration does not raise the passage score above simple sentence averaging, the method is not actually capturing long-range contradictions.

Watch

Extended reading notes

Core claim

The central claim is that hallucination is a relational phenomenon, not a token-level one, and that uncertainty scores should therefore be computed over a semantic graph rather than averaged over independent tokens. The paper's pipeline parses each sentence into an AMR graph, links the sentence graphs into a passage graph via coreference resolution and entity linking, and identifies the entity tokens and relation paths inside that graph. On top of this graph, relation-based uncertainty propagation sends an object entity's uncertainty from its subject entities along dependency edges, weighting by attention and discounting by relation intensity to avoid the overestimation that plagues propagation over all preceding tokens. For passage-level scoring, graph-based uncertainty calibration multiplies each sentence's uncertainty by the contradiction probability a natural-language-inference model assigns between that sentence and each of its graph neighbors. The authors' claim is that these two graph mechanisms, propagation and calibration, are what let the method beat token-averaging baselines, especially on moderate or no-hallucination sentences and on passages where separate sentences contradict each other.

Load-bearing premise

The method assumes the automatically built semantic graph is accurate enough that uncertainty flows along true relations and not spurious ones; the authors' manual review of extracted triples in Appendix A.1 shows the graph construction is not fully automatic, so noisy graphs could misdirect propagation and calibration.

Editorial extensions

If this is right

  • Sentence-level detection improves because uncertainty propagates only along real dependency relations, preventing the overestimation that propagation across all preceding tokens suffers on moderate or no-hallucination sentences.
  • Passage-level detection improves because a sentence that contradicts a non-adjacent neighbor in the semantic graph receives a boosted uncertainty score, catching hallucinations that sentence averaging would miss.
  • The method is reference-free and single-pass: it needs one forward pass of the LLM for token probabilities plus graph construction and NLI scoring, with no external retrieval or repeated sampling.
  • The same pipeline transfers to a new domain and language: the authors report gains on both English WikiBio and Chinese NoteSum, and ablations attribute the improvement to the graph components rather than to token-level statistics alone.

Reading between the lines

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

  • Editorial inference: because triple extraction required manual review in the paper's own setup, a robustness check on fully automatic graphs, without the manual cleanup documented in Appendix A.1, would show whether the approach scales beyond these two datasets.
  • Editorial inference: the contradiction-calibration step could be turned into an explanation tool; the specific neighbor sentence and relation path that drive a high calibrated score are readable evidence for why a passage was flagged, which the paper does not develop.
  • Editorial inference: a natural extension is to replace or augment the NLI model with external fact sources, since the graph already identifies the entities and relations whose factual accuracy would need checking; the paper only mentions this possibility as future work.
  • Editorial inference: the method's reliance on AMR parsing suggests it may be weaker on informal or heavily elliptical text, where AMR parses are noisier; testing on dialogue or social-media data would bound the method's domain of validity.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes UncerSema4HalluDetec, an uncertainty-based hallucination detection method that constructs a passage-level semantic graph from AMR parses, coreference resolution, and entity linking; propagates entity-level uncertainty along graph edges via relation-based propagation (Eq. 2); and calibrates passage-level uncertainty with NLI contradiction probabilities over neighboring sentences (Eq. 7). The method is evaluated on the WikiBio dataset and on a new Chinese NoteSum dataset, reporting improvements over GPT-3 uncertainty baselines, SelfCheckGPT, and FOCUS, with a headline 19.78% improvement in passage-level Spearman correlation.

Significance. If the evaluation were valid, the idea of using a semantic graph to propagate and calibrate token-level uncertainty is a plausible and potentially useful contribution to single-pass hallucination detection. The paper is also transparent in reporting full hyperparameter sweeps and substitution experiments. However, the current experimental protocol does not support the claimed gains: the semantic graph is manually corrected on the test passages, and the hyperparameters and projection functions are selected on the WikiBio test set. The reported improvements are therefore not attributable to the automatic algorithm as described in the abstract.

major comments (4)
  1. [Appendix A.1; Section 'Semantic Graph Construction'; Table 2] The evaluation uses manually corrected semantic graphs on the test passages. Appendix A.1 states: 'Finally, we manually review all the extracted triples to obtain more accurate results.' Since the graph edges directly enter the propagated uncertainty (Eq. 2) and the neighbor set used for calibration (Eq. 7), every result in Table 2—including the 19.78% passage-level gain—is produced by a human-assisted pipeline, not by the automatic AMR/spaCy pipeline described in the main text. The paper neither reports results with fully automatic graph construction nor quantifies the effect of the manual review. Consequently, the abstract's characterization of the method as reference-free and efficient is not supported, and the headline numbers cannot be reproduced without human labor on each test passage.
  2. [Appendix A.5; Tables 4-8; Table 2] The free parameters α, β, λ, k and the sentence- and passage-level projection functions are chosen by maximizing the reported evaluation metrics on the WikiBio dataset with LLaMA-30B. Appendix A.5 says 'we adjust the following four parameters ... on the WikiBio dataset' and 'we test on three projection functions ... on the WikiBio dataset.' Because no separate validation set is described, the WikiBio numbers in Table 2 are selection results rather than an unbiased estimate of performance. The reported improvements and the ablation conclusions in Table 3 therefore do not establish generalization; the comparisons with baselines are also affected because the baseline scores are not tuned in the same way.
  3. [Appendix A.4; Table 2] The NoteSum comparison is not controlled for backbone. Appendix A.4 states that for the Chinese dataset the FOCUS baseline uses Alpaca-Chinese-13B and Alpaca-Chinese-33B, while Table 2 labels the proposed method's rows as LLaMA-13B and LLaMA-30B. If the proposed method was run with the original LLaMA models on Chinese text while the baseline used a Chinese-tuned model, the comparison is confounded; if the proposed method also used the Chinese-adapted models, this is not stated. The paper should specify the exact backbone used for the proposed method on NoteSum and run baselines under identical conditions.
  4. [Section 'Passage-level Uncertainty', Eq. (7); Table 3] The contribution of the graph-based calibration is not isolated from the manual graph correction. The '- graph' ablation in Table 3 compares the full model against averaging sentence uncertainties, but both conditions use the manually reviewed triples from Appendix A.1. Therefore the drop of about 2 Pearson points cannot be attributed to the automatic semantic graph; it may reflect the human-corrected neighbor structure. A proper comparison would use the same automatic graph construction in both conditions and report the effect of manual correction separately.
minor comments (5)
  1. [Eq. (2)] The attention score att(s', o) in Eq. (2) does not specify which layer or head of the LLaMA attention matrix is used; since attention scores are not probabilities over semantic relations, this choice should be stated for reproducibility.
  2. [Eq. (1)] The notation C_j^i is introduced as a sorted list of top-k probabilities, but the formula applies max() and variance to it; defining C_j^i as the set of top-k probability values would remove the ambiguity.
  3. [Appendix A.2] The annotation agreement of 0.76 for NoteSum is reported without defining the metric; please specify whether this is Cohen's kappa, Krippendorff's alpha, or simple percentage agreement.
  4. [Appendix A.4] The baseline description contains an incomplete citation with a placeholder '( ?)' after 'ChatGPT3.5'; this should be corrected.
  5. [Table 6] The λ sweep is non-monotonic (e.g., NonFact* is 65.34 at λ=0.6 and 61.16 at λ=0.7), yet the chosen value is justified only by the tabulated point estimates; reporting variance or confidence intervals across runs would strengthen the sensitivity analysis.

Circularity Check

1 steps flagged · score 6.0 of 10

WikiBio results reuse the hyperparameter/projection search grid as the reported test numbers; NoteSum remains an independent check.

  1. fitted input called prediction [Appendix A.5 (Prompts, Hyper-parameters and Projection Functions), Tables 4-8; compared against Table 2 Main Results.]
    "We adjust the following four parameters: α, β, λ, and k on the WikiBio dataset with LLaMA-30B, and the results are shown in Table 4, Table 5, Table 6 and Table 7, respectively. When adjusting the parameters, we ensure that the other parameters remain fixed. According to the results in the tables, α, β, λ, and k are set as 0.8, 0.65, 0.7, and 3 respectively. ... According to the results in the table, we choose a logistic function for sentences and an inverse function for passages."

    The OURS/LLaMA-30B row in Table 2 on WikiBio is exactly the configuration selected by maximizing the same Table 2 metrics on the same WikiBio test set. For instance, Table 4's α=0.8 row reports NonFact* AUC 61.16, Pearson 77.60, Spearman 74.44, and Table 2's WikiBio OURS row reports the identical values; the same reuse holds for the other tuned parameters and the two projection functions. Thus the WikiBio sentence/passage scores are not an out-of-sample test of the method; they are the selection criterion being reported as the result. The NoteSum results are less affected because the parameters were fixed before evaluation there, but the paper's blanket claim of consistent improvement on WikiBio is partly forced by this construction.

full rationale

The paper's core derivation—token uncertainty (Eq. 1), relation-based propagation (Eqs. 2-4), quantile global uncertainty (Eq. 5), interpolation (Eq. 6), and graph-calibrated passage score (Eq. 7)—is a coherent, self-contained scoring model; no equation reduces to the hallucination labels by definition, and no load-bearing claim rests on a self-citation. The main circularity is the Appendix A.5 test-set tuning: α, β, λ, k and the sentence/passage projection functions are chosen by inspecting WikiBio metrics, and the same numbers then appear as the WikiBio results in Table 2, so that portion of the empirical claim is an in-sample selection artifact. Additionally, Appendix A.1 states 'we manually review all the extracted triples to obtain more accurate results'; if this review was applied to the evaluation passages, Table 2 reflects human-corrected semantic graphs rather than the automatic AMR/spaCy pipeline, a serious validity concern though not a definitional circularity because there is no evidence the review used hallucination labels. These issues do not make the NoteSum comparison circular, since NoteSum was not used for parameter selection, but they do prevent the WikiBio numbers from being treated as an independent prediction.

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

The method rests on several modeling choices that are not derived from first principles: the sequence decay term, the relation intensity penalty, the quantile-based global uncertainty, and the NLI calibration. The most consequential are the four hyperparameters tuned on the evaluation set and the manual review of graph triples, both documented in the appendix.

free parameters (5)
  • alpha (quantile level) = 0.8
    Selected by sweeping on WikiBio with LLaMA-30B (Table 4); controls the quantile in Equation 5.
  • beta (propagation weight) = 0.65
    Selected by sweeping on WikiBio (Table 5); balances self and propagated entity uncertainty in Equation 4.
  • lambda (interpolation weight) = 0.7
    Selected by sweeping on WikiBio (Table 6); mixes entity and global uncertainty in Equation 6.
  • k (top-k probability cutoff) = 3
    Selected by sweeping on WikiBio (Table 7); defines the candidate token set in Equation 1.
  • Projection functions for sentence and passage scores = logistic for sentence, inverse for passage
    Chosen based on Table 8 performance on the WikiBio evaluation set; no held-out validation described.
assumptions (6)
  • domain assumption AMR parsing produces accurate semantic graphs for each sentence.
    The method relies on the AMR parser (Xu et al. 2023) to extract triples; parsing errors propagate into uncertainty computation.
  • domain assumption Coreference resolution and entity linking correctly connect entities across sentences.
    The passage-level graph depends on spaCy-based linking; noisy linking would corrupt neighbor relations for calibration.
  • domain assumption Attention scores between tokens are meaningful semantic relatedness weights.
    Equations 2 and 3 use attention scores between subject, relation, and object tokens as propagation weights, but the paper does not specify which layer or head is used.
  • domain assumption NLI contradiction probability reliably measures sentence conflict.
    Equation 7 uses NLI(con|Sj,Si) as a calibration weight; unreliable NLI would misdirect passage-level scores.
  • ad hoc to paper Hallucination propagates along semantic graph edges and relation intensity indicates factuality confidence.
    This is the paper's central modeling assumption, stated without independent evidence: 'high relation intensities usually indicate high factuality-confidence' (Section Sentence-level Uncertainty).
  • domain assumption Hallucination accumulates with sequence length.
    Used in the sequence decay term in Equation 1, based on prior studies cited by the authors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Uncertainty Modeling with Semantic Graph for Hallucination Detection." pith.science (2026). https://pith.science/paper/EFZKSEAQ

@misc{pith2026250102020,
  author       = {Pith},
  title        = {Pith review of: Enhancing Uncertainty Modeling with Semantic Graph for Hallucination Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EFZKSEAQ}},
  note         = {Machine review of arXiv:2501.02020}
}
read the original abstract

Large Language Models (LLMs) are prone to hallucination with non-factual or unfaithful statements, which undermines the applications in real-world scenarios. Recent researches focus on uncertainty-based hallucination detection, which utilizes the output probability of LLMs for uncertainty calculation and does not rely on external knowledge or frequent sampling from LLMs. Whereas, most approaches merely consider the uncertainty of each independent token, while the intricate semantic relations among tokens and sentences are not well studied, which limits the detection of hallucination that spans over multiple tokens and sentences in the passage. In this paper, we propose a method to enhance uncertainty modeling with semantic graph for hallucination detection. Specifically, we first construct a semantic graph that well captures the relations among entity tokens and sentences. Then, we incorporate the relations between two entities for uncertainty propagation to enhance sentence-level hallucination detection. Given that hallucination occurs due to the conflict between sentences, we further present a graph-based uncertainty calibration method that integrates the contradiction probability of the sentence with its neighbors in the semantic graph for uncertainty calculation. Extensive experiments on two datasets show the great advantages of our proposed approach. In particular, we obtain substantial improvements with 19.78% in passage-level hallucination detection.

Figures

Figures reproduced from arXiv: 2501.02020 by the authors.

Figure 1
Figure 1. (a) Previous works only concern independent to [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overview of our approach for hallucination detection. For token-level uncertainty, we integrate the maximum [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 5
Figure 5. The Pearson and Spearman metrics of ours and the [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization of the entity uncertainty and global [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 33 canonical work pages

  1. [1]

    Aracena, G.; Luster, K.; Santos, F.; Steinmacher, I.; and Gerosa, M. A. 2024. Applying Large Language Models API to Issue Classification Problem. CoRR, abs/2401.04637

    work page Pith review arXiv 2024

  2. [2]

    Bradley, A. P. 1997. The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognit., 30(7): 1145--1159

    work page 1997

  3. [3]

    Chen, J.; Lin, H.; Han, X.; and Sun, L. 2024 a . Benchmarking Large Language Models in Retrieval-Augmented Generation. In AAAI 2024, 17754--17762. AAAI Press

    work page 2024

  4. [4]

    Chen, K.; Chen, Q.; Zhou, J.; He, Y.; and He, L. 2024 b . DiaHalu: A Dialogue-level Hallucination Evaluation Benchmark for Large Language Models. CoRR, abs/2403.00896

    arXiv 2024

  5. [5]

    Chen, K.; Zhou, J.; Chen, Q.; Liu, S.; and He, L. 2024 c . A Regularization-based Transfer Learning Method for Information Extraction via Instructed Graph Decoder. In LREC/COLING 2024, 1472--1485. ELRA and ICCL

    work page 2024

  6. [6]

    Chen, L.; Deng, Y.; Bian, Y.; Qin, Z.; and et al. 2023. Beyond Factuality: A Comprehensive Evaluation of Large Language Models as Knowledge Generators. In EMNLP 2023, 6325--6341. Association for Computational Linguistics

    work page 2023

  7. [7]

    X.; He, L.; and An, W

    Chen, Q.; Hu, Q.; Huang, J. X.; He, L.; and An, W. 2017. Enhancing Recurrent Neural Networks with Positional Attention for Question Answering. In SIGIR, 2017, 993--996. ACM

    work page 2017

  8. [8]

    Choi, S.; Fang, T.; Wang, Z.; and Song, Y. 2023. KCTS: Knowledge-Constrained Tree Search Decoding with Token-Level Hallucination Detection. In EMNLP 2023, 14035--14053. Association for Computational Linguistics

    work page 2023

Show all 49 references
  1. [9]

    Cohen, I.; Huang, Y.; Chen, J.; Benesty, J.; Benesty, J.; Chen, J.; Huang, Y.; and Cohen, I. 2009. Pearson correlation coefficient. Noise reduction in speech processing, 1--4

    2009

  2. [10]

    Cui, J.; Li, Z.; Yan, Y.; Chen, B.; and Yuan, L. 2023. ChatLaw: Open-Source Legal Large Language Model with Integrated External Knowledge Bases. CoRR, abs/2306.16092

    2023 arXiv

  3. [11]

    Dong, Q.; Liu, Y.; Ai, Q.; Wu, Z.; and et al. 2024. Unsupervised Large Language Model Alignment for Information Retrieval via Contrastive Feedback. In SIGIR 2024, 48--58. ACM

    2024

  4. [12]

    Floridi, L.; and Chiriatti, M. 2020. GPT-3: Its Nature, Scope, Limits, and Consequences. Minds Mach., 30(4): 681--694

    2020

  5. [13]

    Giulianelli, M.; Baan, J.; Aziz, W.; Fern \' a ndez, R.; and Plank, B. 2023. What Comes Next? Evaluating Uncertainty in Neural Text Generators Against Human Production Variability. In EMNLP 2023, 14349--14371. Association for Computational Linguistics

    2023

  6. [14]

    S.; Menon, A

    Gupta, N.; Narasimhan, H.; Jitkrittum, W.; Rawat, A. S.; Menon, A. K.; and Kumar, S. 2024. Language Model Cascades: Token-level uncertainty and beyond. CoRR, abs/2404.10136

    2024 arXiv

  7. [15]

    He, P.; Gao, J.; and Chen, W. 2023. DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing. In ICLR 2023. OpenReview.net

    2023

  8. [16]

    Huang, L.; Yu, W.; Ma, W.; Zhong, W.; Feng, Z.; and et al. 2023 a . A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions. CoRR, abs/2311.05232

    2023 arXiv

  9. [17]

    Huang, Y.; Song, J.; Wang, Z.; Chen, H.; and Ma, L. 2023 b . Look Before You Leap: An Exploratory Study of Uncertainty Measurement for Large Language Models. CoRR, abs/2307.10236

    2023 arXiv

  10. [18]

    Kryscinski, W.; McCann, B.; Xiong, C.; and Socher, R. 2020. Evaluating the Factual Consistency of Abstractive Text Summarization. In EMNLP 2020, 9332--9346. Association for Computational Linguistics

    2020

  11. [19]

    Lai, H.; and Nissim, M. 2024. A Survey on Automatic Generation of Figurative Language: From Rule-based Systems to Large Language Models. ACM Comput. Surv. , 56(10): 244

    2024

  12. [20]

    C.; and Song, M

    Lee, J.; Stevens, N.; Han, S. C.; and Song, M. 2024. A Survey of Large Language Models in Finance (FinLLMs). CoRR, abs/2402.02315

    2024 arXiv

  13. [21]

    Malkin, N.; Wang, Z.; and Jojic, N. 2022. Coherence boosting: When your pretrained language model is not paying enough attention. In ACL 2022, 8214--8236. Association for Computational Linguistics

    2022

  14. [22]

    Manakul, P.; Liusie, A.; and Gales, M. J. F. 2023. SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models. In EMNLP 2023, 9004--9017. Association for Computational Linguistics

    2023

  15. [23]

    Mündler, N.; He, J.; Jenko, S.; and Vechev, M. 2024. Self-contradictory Hallucinations of Large Language Models: Evaluation, Detection and Mitigation. arXiv:2305.15852

    2024 arXiv

  16. [24]

    Nasz \' a di, K.; Manggala, P.; and Monz, C. 2023. Aligning Predictive Uncertainty with Clarification Questions in Grounded Dialog. In Findings of EMNLP 2023, 14988--14998. Association for Computational Linguistics

    2023

  17. [25]

    Pagnoni, A.; Balachandran, V.; and Tsvetkov, Y. 2021. Understanding Factuality in Abstractive Summarization with FRANK: A Benchmark for Factuality Metrics. In NAACL-HLT 2021, 4812--4829. Association for Computational Linguistics

    2021

  18. [26]

    Petersen, F.; Mishra, A.; Kuehne, H.; Borgelt, C.; Deussen, O.; and Yurochkin, M. 2024. Uncertainty Quantification via Stable Distribution Propagation. arXiv:2402.08324

    2024 arXiv

  19. [27]

    Sedgwick, P. 2014. Spearman’s rank correlation coefficient. Bmj, 349

    2014

  20. [28]

    Sheng, Z.; Zhang, T.; Jiang, C.; and Kang, D. 2024. BBScore: A Brownian Bridge Based Metric for Assessing Text Coherence. In AAAI 2024, 14937--14945. AAAI Press

    2024

  21. [29]

    Siino, M. 2024. B rain L lama at S em E val-2024 Task 6: Prompting Llama to detect hallucinations and related observable overgeneration mistakes. In Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024), 82--87. Mexico City, Mexico: Association f...

    2024

  22. [30]

    Varshney, N.; Yao, W.; Zhang, H.; Chen, J.; and Yu, D. 2023. A Stitch in Time Saves Nine: Detecting and Mitigating Hallucinations of LLMs by Validating Low-Confidence Generation. CoRR, abs/2307.03987

    2023 arXiv

  23. [31]

    Wang, C.; Liu, X.; Yue, Y.; Tang, X.; Zhang, T.; and et al. 2023 a . Survey on Factuality in Large Language Models: Knowledge, Retrieval and Domain-Specificity. CoRR, abs/2310.07521

    2023 arXiv

  24. [32]

    Wang, J.; Sun, Q.; Chen, N.; Wang, C.; Huang, J.; Gao, M.; and Li, X. 2023 b . Uncertainty-aware Parameter-Efficient Self-training for Semi-supervised Language Understanding. arXiv:2310.13022

    2023 arXiv

  25. [33]

    Wang, X.; Yan, Y.; Huang, L.; Zheng, X.; and Huang, X. 2023 c . Hallucination Detection for Generative Large Language Models by Bayesian Sequential Estimation. In EMNLP 2023, Singapore, December 6-10, 2023, 15361--15371. Association for Computational Linguistics

    2023

  26. [34]

    Xiong, M.; Hu, Z.; Lu, X.; Li, Y.; and et al. 2023. Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs. CoRR, abs/2306.13063

    2023 arXiv

  27. [35]

    Xiong, M.; Hu, Z.; Lu, X.; LI, Y.; Fu, J.; He, J.; and Hooi, B. 2024. Can LLM s Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLM s. In The Twelfth International Conference on Learning Representations

    2024

  28. [36]

    Y.; and Huang, L

    Xu, Z.; Lee, J. Y.; and Huang, L. 2023. Learning from a Friend: Improving Event Extraction via Self-Training with Feedback from A bstract M eaning R epresentation. In Rogers, A.; Boyd-Graber, J.; and Okazaki, N., eds., Findings of the Association for Computational Linguistics:...

    2023

  29. [37]

    Yan, L.; Sha, L.; Zhao, L.; Li, Y.; and et al. 2024. Practical and ethical challenges of large language models in education: A systematic scoping review. Br. J. Educ. Technol., 55(1): 90--112

    2024

  30. [38]

    Ye, L.; Lei, Z.; Yin, J.; Chen, Q.; Zhou, J.; and He, L. 2024. Boosting Conversational Question Answering with Fine-Grained Retrieval-Augmentation and Self-Check. In SIGIR 2024, 2301--2305. ACM

    2024

  31. [39]

    Zhang, D.; Yu, Y.; Li, C.; Dong, J.; and et al. 2024. MM-LLMs: Recent Advances in MultiModal Large Language Models. CoRR, abs/2401.13601

    2024 arXiv

  32. [40]

    A.; and Sricharan, K

    Zhang, J.; Li, Z.; Das, K.; Malin, B. A.; and Sricharan, K. 2023 a . SAC \( ^3 \) : Reliable Hallucination Detection in Black-Box Language Models via Semantic-aware Cross-check Consistency. In Findings of EMNLP 2023, 15445--15458. Association for Computational Linguistics

    2023

  33. [41]

    Zhang, J.; Muhamed, A.; Anantharaman, A.; Wang, G.; Chen, C.; and et al. 2023 b . ReAugKD: Retrieval-Augmented Knowledge Distillation For Pre-trained Language Models. In ACL 2023, 1128--1136. Association for Computational Linguistics

    2023

  34. [42]

    Zhang, T.; Qiu, L.; Guo, Q.; Deng, C.; and et al. 2023 c . Enhancing Uncertainty-Based Hallucination Detection with Stronger Focus. In EMNLP 2023, 915--932. Association for Computational Linguistics

    2023

  35. [43]

    Zhao, L.; Nguyen, K.; and Daume, H. 2023. Hallucination Detection for Grounded Instruction Generation. In Findings of EMNLP 2023, 4044--4053. Association for Computational Linguistics

    2023

  36. [44]

    X.; Zhou, K.; Li, J.; Tang, T.; Wang, X.; Hou, Y.; and et al

    Zhao, W. X.; Zhou, K.; Li, J.; Tang, T.; Wang, X.; Hou, Y.; and et al. 2023 a . A Survey of Large Language Models. CoRR, abs/2303.18223

    2023 arXiv

  37. [45]

    Zhao, Y.; Yan, L.; Sun, W.; Xing, G.; and et al. 2023 b . Knowing What LLMs DO NOT Know: A Simple Yet Effective Self-Detection Method. CoRR, abs/2310.17918

    2023 arXiv

  38. [46]

    Zheng, Z.; and Zhu, X. 2023. NatLogAttack: A Framework for Attacking Natural Language Inference Models with Natural Logic. In ACL 2023, 9960--9976. Association for Computational Linguistics

    2023

  39. [47]

    Zhou, Y.; Hu, H.; Yu, J.; Xu, Z.; Lu, W.; and Cao, Y. 2023. A Solution to Co-occurrence Bias: Attributes Disentanglement via Mutual Information Minimization for Pedestrian Attribute Recognition. CoRR, abs/2307.15252

    2023 arXiv

  40. [48]

    , " * 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...

  41. [49]

    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 10, 2026 · model on record in the stance chip above.