Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models

Chaozhuo Li; Chenxu Wang; Feiran Huang; Litian Zhang; Pengbo Wang; Philip S. Yu; Qiwei Ye; Xi Zhang; Yuanbo Xu; Zheng Liu

REVIEW 4 major objections 5 minor 2 cited by

Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This survey argues that hallucination in large language models is not a fixable bug but a mathematical necessity rooted in undecidability.

desk verdict The survey half is useful reference material; the mathematical core claiming hallucination is unavoidable does not hold up as written. read the letter →

arxiv 2507.02870 v1 pith:QQ6AFM4J submitted 2025-06-06 cs.CL

Chaozhuo Li , Pengbo Wang , Chenxu Wang , Litian Zhang , Zheng Liu , Qiwei Ye , Yuanbo Xu , Feiran Huang

show 2 more authors

Xi Zhang Philip S. Yu

This is my paper

classification cs.CL

keywords largelanguagemodelshallucinationundecidabilityGödelincompletenesshaltingproblemacceptancefaithfulnesslogicalinconsistency

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This survey tries to establish that hallucination in large language models is not a fixable bug but a mathematical necessity. It defines hallucination formally as any generated response outside the canonical response set $F(q)$ for a query $q$, splits the phenomenon into factual, faithfulness, and logical-inconsistency types, and then argues that each type traces back to undecidability results: Gödel's incompleteness theorems, the halting problem, and the acceptance problem. If the argument is right, no amount of better data, larger architectures, or alignment training can fully eliminate hallucinations; mitigation can only reduce their frequency and impact. The survey also organizes current detection, evaluation, and mitigation methods and explains why each is constrained by the same underlying limits.

What carries the argument

The argument runs on undecidability transfers. The load-bearing objects are the formal definition of hallucination as $LLM(q)\notin F(q)$; the ideal LLM Turing machine $T$ whose perfect intent classification would imply a decider $D_T$; the reduction from the acceptance problem $L_a$ to that decider; Gödel's second incompleteness theorem, written $S\not\Rightarrow\mathrm{Con}(S)$, applied to the LLM as a formal system; the halting problem's undecidability applied to token generation and stopping; and the sequential conditional-probability decomposition $P(w)=P(w_1)P(w_2\mid w_1)\cdots P(w_n\mid w_1,\ldots,w_{n-1})$, which shows the model commits to each token without global foresight. Each of these carries one part of the conclusion that factual, faithfulness, and logical hallucinations are unavoidable.

What would settle it

A concrete test: restrict an LLM to a decidable fragment of arithmetic or propositional-logic entailment, attach a machine-checkable proof verifier, and require the system to emit only outputs that pass the verifier. If such a system can be built and reaches zero emitted hallucinations over a large sample, the claim that hallucinations are unavoidable for every sufficiently capable system would need to be restricted; if the raw model still emits non-canonical outputs on decidable queries, that supports the paper's position.

Watch

Extended reading notes

Core claim

The paper's central claim is that the underlying cause of hallucinations is not tied to specific design features of LLMs or the limitations of their datasets but stems from deeper mathematical and logical issues inherent to undecidability. To support this, it constructs an idealized LLM as a Turing machine $T$ with perfect intent classification and argues that such a $T$ would yield a decider $D_T$ for whether an input pair $\langle T, w\rangle$ has the correct intent; from $D_T$ it builds a decider for the acceptance problem $L_a=\{\langle M,w\rangle: M \text{ accepts } w\}$, which is known to be undecidable, producing a contradiction. It applies Gödel's second incompleteness theorem ($S \not\Rightarrow \mathrm{Con}(S)$) to argue that an LLM cannot verify the consistency of its own generation, and it cites the infinite size of $\Sigma^*$ to argue that exhaustive search over possible outputs is impossible, so the model can never know the globally optimal or fully consistent completion in advance.

Load-bearing premise

The load-bearing premise is that a trained LLM can be treated as a formal system of the kind Gödel's theorems and the halting problem apply to, and that a perfect intent classifier would really give a decider for the acceptance problem; the paper asserts these steps rather than proving the representability of arithmetic inside an LLM or the validity of the reduction.

Editorial extensions

If this is right

Factual hallucinations cannot be fully eliminated: any sufficiently expressive LLM will generate statements that are unverifiable within its own formal system.
Faithfulness hallucinations are unavoidable because perfect intent classification would solve the acceptance problem, which is undecidable.
Logical consistency cannot be guaranteed, since the model cannot prove its own consistency and cannot foresee the global structure of its output during sequential token generation.
Existing mitigation strategies, including retrieval augmentation, knowledge graphs, prompt engineering, fine-tuning, and decoding changes, can reduce hallucination frequency and severity but cannot make generation hallucination-free.
Detection and evaluation should therefore aim at calibrated uncertainty and safe abstention rather than perfect correctness.

Reading between the lines

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

If the inevitability claim holds, the practical target for deployed systems should shift from eliminating hallucinations to containing them through abstention, external verification, and explicit uncertainty signaling.
A testable consequence of the paper's position is that scaling data and parameters will show diminishing returns on hallucination rates, with a residual floor set by undecidable statements.
The paper's 'truthful subspace' proposal suggests a research program: if truthfulness has a geometric representation in activation space, interventions could steer models toward it, but the paper's own argument implies such steering can reduce, not remove, hallucinations.
The same undecidability reasoning would extend to multimodal and agentic LLMs, where 'faithfulness' also covers tool use and action constraints; the intent-classification obstacle would apply there as well.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The part of this paper worth keeping is the survey: the taxonomy of factual, faithfulness, and logical hallucinations, the catalog of detection methods, and the mitigation strategies are organized sensibly and cover recent work. A reader new to the area could use Sections II and IV as a starting point. That is real, if modest, value.

The advertised original contribution is the inevitability argument in Section III, and that argument is not sound. The paper claims LLMs inherit Gödel incompleteness and the undecidability of the halting problem, but it never establishes that an LLM is a formal system with an axiom system, proof relation, and arithmetic representability. The reduction in III.B is the load-bearing step and it fails: the constructed decider D_T is about intent classification, not about whether an arbitrary machine M accepts w. No map from acceptance instances to intent-classification instances is given, so the contradiction with undecidability is not derived. The extra step about needing to explore all of Σ* conflates sampling from a conditional distribution with exhaustive search. I agree with the stress-test note: the central claim does not follow.

There are more concrete problems. The paper promises a "task-aware evaluation taxonomy" as a contribution, but I could not find it anywhere in the text — that is a missing deliverable. The bibliography contains a placeholder reference [23] with authors "A. Author and B. Coauthor" and a fake JMLR page, which is unacceptable in any draft. Some of the speculative concepts (truthful subspace, hallucination space) are presented as future directions, which is fine, but they are not new results.

To be fair, the paper does cite the existing inevitability arguments (Banerjee et al., Xu et al., Kalai and Vempala), and it mostly summarizes them accurately. The problem is that it adds a faulty reformulation and presents it as a proof. If the authors removed the invalid reduction and the Gödel analogy, and were honest that the survey is a synthesis rather than a new theoretical result, the survey would be serviceable. As is, the central argument is misleading.

The reader's reject verdict is proportionate. I would not send this to peer review in its current form; a serious referee would spend most of their time on the flawed Section III and the placeholder reference. If the authors revise to cut the false inevitability claim and clean up the bibliography, it could become a usable survey, but that is not what is in front of us now.

Referee Report

4 major / 5 minor

Summary. The paper is a survey of hallucination in large language models. It proposes a formal definition of hallucination, a three-part taxonomy (factual, faithfulness, logical), and then advances the central claim that hallucinations are mathematically unavoidable. Section III derives this inevitability from Gödel's incompleteness theorems, the undecidability of the acceptance problem, and the infinitude of the string space. The remainder of the paper catalogues evaluation metrics, benchmarks, detection methods, and mitigation strategies, and closes with limitations and future directions.

Significance. If the Section III claim were established, it would be a significant negative result: it would imply that factual, faithfulness, and logical hallucinations cannot be fully eliminated by any data, architecture, or alignment strategy. The survey also offers a useful structured synthesis of the detection and mitigation literature and a comparative table of prior surveys. However, the mathematical contribution is not accompanied by machine-checked proofs, reproducible code, or a parameter-free derivation; it rests on an informal analogy to computability theory. The review sections are a reasonable empirical resource, but the advertised theoretical foundation is the paper's central contribution and is not sound.

major comments (4)

[III.A and III.C] Gödel's incompleteness theorems apply to formal systems with an axiomatized proof relation and the ability to represent arithmetic. The paper asserts in Section III.A that 'LLMs, particularly those based on Transformer architectures, possess significant expressive power' and concludes 'according to Gödel's theorem ... there will always be statements generated by the LLM that cannot be verified,' and in Section III.C that 'LLMs, as complex formal systems, inherit this limitation.' No axiom system, proof relation, or arithmetic representation is specified for an LLM, and the generation process is a conditional probability distribution rather than a deductive system. The conclusion that factual hallucinations and the inability to self-verify are unavoidable therefore does not follow from the cited theorems.
[III.B] The reduction to the acceptance problem is invalid. The assumed decider D_T is defined over whether an input pair ⟨T,w⟩ 'corresponds to the correct intent,' which is an intent-classification decision; the acceptance problem L_a = {⟨M,w⟩: M accepts w} asks whether an arbitrary machine M accepts w. The constructed D_M never uses M, and no mapping from ⟨M,w⟩ to an intent-classification instance is provided. Consequently, 'D_T accepts' is not equivalent to 'M accepts w,' and no contradiction with the undecidability of the acceptance problem is established. The paragraph's conclusion that 'perfect intent classification is, in principle, impossible' is unsupported.
[III.B, third prerequisite] The argument that the model cannot select an optimal response because, with |Σ*|=∞, it would need to 'explore all potential outputs' conflates exhaustive search over a string space with autoregressive sampling from a conditional distribution. An LLM does not enumerate Σ*; it conditions on its prefix and samples or searches a finite set of candidate continuations at each step. The infinitude of Σ* therefore does not by itself make response selection infeasible, and the claimed unattainability of the third prerequisite does not follow.
[III.C] The use of Gödel's second incompleteness theorem repeats the unsupported inheritance step: writing 'LLM̸⇒Con(LLM)' only makes sense if the LLM is shown to be a formal system that can express its own consistency, which the paper does not do. The additional claim that the undecidability of the halting problem means an LLM 'cannot determine in advance when it should stop generating' also conflates a mathematical property of a program with an engineering decision; in practice autoregressive models terminate by emitting an end-of-sequence token, not by solving the halting problem. The conclusion that logical inconsistencies are 'fundamentally unavoidable' is therefore not established by the material in this subsection.

minor comments (5)

[References] Reference [23] is a placeholder: the authors are listed as 'A. Author and B. Coauthor' and the URL 'jmlr.org/papers/v25/author2024.html' does not resolve to a real paper. This citation-integrity problem must be fixed before resubmission.
[Author affiliations] The affiliation line reads 'Chaozhuo Li, Pengbo Wang, Chuanshi and Xi Zhang are with Beijing University of Posts and Telecommunications,' but no author named Chuanshi appears in the author list; this appears to be a typo that should be corrected.
[Figures 6 and 7] The captions of Figures 6 and 7 appear to be swapped: Figure 6 is labeled 'A Standard Paradigm of Shifting Demand Strategy' but depicts RAG, Knowledge Graphs, Prompt Engineering, and Reflection, which the text describes under Task Simplification Strategy, while Figure 7 is labeled 'A Standard Paradigm of Shifting Demand Strategy' again but depicts fine-tuning, attention, activation, and decoding, which belong to Capability Enhancement Strategy.
[Figure 5 caption] The caption of Figure 5 contains the typo 'Tast Simplification Strategy' and should read 'Task Simplification Strategy.'
[IV.C.1] The text contains 'gekhman et al. [147]' with a lowercase name; the reference should be styled as 'Gekhman et al.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the inevitability claim rests on external undecidability theorems; the Section III.B reduction is invalid on other grounds, not circular.

full rationale

The paper's central claim—that factual, faithfulness, and logical hallucinations are unavoidable—is argued from Gödel's incompleteness theorems and the undecidability of the halting problem, both cited as external mathematical results with assumptions that do not include the paper's conclusion. No parameter is fitted to data and then relabeled as a prediction; no equation in Section II is reused by construction to manufacture Section III's conclusion. The few self-citations (e.g., [149] on culturally nuanced questions) are contextual and not load-bearing for the inevitability thesis. There is a serious non-circular validity problem: in Section III.B, the purported decider D_M simply echoes D_T and never references the machine M from the acceptance problem, so no contradiction with L_a is actually derived; and the assertion that LLMs are 'complex formal systems' inheriting Gödelian limits is an analogy, not a proof. These are soundness and correctness concerns, not cases where the result is equivalent to its inputs by construction. Therefore, no circularity is present.

Assumptions & free parameters 0 free parameters · 6 assumptions · 3 invented entities

No free parameters or fitted values appear in this survey; the central argument depends on applying standard undecidability theorems to LLMs, plus several unproved domain assumptions about how LLMs relate to formal systems.

assumptions (6)

standard math Gödel's first incompleteness theorem: any consistent formal system rich enough for arithmetic contains unprovable truths.
Invoked in Section III.A to argue LLM outputs are inevitably unverifiable; the theorem itself is standard, but its applicability to LLMs is the unproved step.
standard math Gödel's second incompleteness theorem: a consistent system rich enough for arithmetic cannot prove its own consistency.
Invoked in Section III.C to conclude LLMs cannot self-validate consistency.
standard math Undecidability of the acceptance problem and the halting problem.
Used in Section III.B to argue perfect intent classification and stopping are impossible for an ideal LLM Turing machine.
ad hoc to paper LLMs are formal systems to which the incompleteness theorems apply.
Sections III.A and III.C assert this without proof; LLMs are trained probabilistic models, not axiomatic systems, so the theorems do not obviously apply.
ad hoc to paper Perfect intent classification would imply a decider for the acceptance problem.
Section III.B constructs decider D_T from an ideal LLM and then defines D_M from D_T; the mapping between <T,w> and <M,w> is not a valid reduction.
ad hoc to paper Selecting an optimal response requires enumerating all strings in Sigma*.
Sections III.B and III.C claim |Sigma*|=infinity makes the third prerequisite unattainable, but autoregressive sampling does not require enumeration of all strings.

invented entities (3)

Truthful subspace and hallucination space in activation space
purpose: Speculative geometric regions suggested in Section V.C for future research on steering LLMs away from hallucination.
No empirical evidence is provided; the paper frames them as open questions, and there is no falsifiable prediction attached.
Super assertions and absolute errors
purpose: Hypothetical categories of statements that resist truth/hallucination classification, proposed in Section V.C.
No operational definition or measurement is given; they are exploratory concepts.
Schrödinger's Memory
purpose: Metaphor describing LLM memory as only existing when invoked, used in Section III.D.3 to explain attention limitations.
Presented as a coined descriptor with no separate evidence beyond the attention-limitation observation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models." pith.science (2026). https://pith.science/paper/QQ6AFM4J

@misc{pith2026250702870,
  author       = {Pith},
  title        = {Pith review of: Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QQ6AFM4J}},
  note         = {Machine review of arXiv:2507.02870}
}

read the original abstract

Edgar Allan Poe noted, "Truth often lurks in the shadow of error," highlighting the deep complexity intrinsic to the interplay between truth and falsehood, notably under conditions of cognitive and informational asymmetry. This dynamic is strikingly evident in large language models (LLMs). Despite their impressive linguistic generation capabilities, LLMs sometimes produce information that appears factually accurate but is, in reality, fabricated, an issue often referred to as 'hallucinations'. The prevalence of these hallucinations can mislead users, affecting their judgments and decisions. In sectors such as finance, law, and healthcare, such misinformation risks causing substantial economic losses, legal disputes, and health risks, with wide-ranging consequences.In our research, we have methodically categorized, analyzed the causes, detection methods, and solutions related to LLM hallucinations. Our efforts have particularly focused on understanding the roots of hallucinations and evaluating the efficacy of current strategies in revealing the underlying logic, thereby paving the way for the development of innovative and potent approaches. By examining why certain measures are effective against hallucinations, our study aims to foster a comprehensive approach to tackling this issue within the domain of LLMs.

Figures

Figures reproduced from arXiv: 2507.02870 by the authors.

**Figure 2.** Overview of the paper workflow [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Formal Representation of Query-Response Mapping in LLMs [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: A Standard Paradigm of Detecting Hallucinations [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

**Figure 5.** Figure 5: Paradigms of Tast Simplification Strategy [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]

**Figure 6.** Figure 6: A Standard Paradigm of Shifting Demand Strategy [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]

**Figure 7.** Figure 7: A Standard Paradigm of Shifting Demand Strategy [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Not All Needles Are Found: How Fact Distribution and Don't Make It Up Prompts Shape Retrieval, Reasoning, and Hallucination in Long-Context LLMs
cs.CL 2026-01 conditional novelty 5.0 of 10

On a new extended needle-in-a-haystack benchmark, explicit anti-hallucination prompts and dispersed fact placement cause some long-context LLMs to over-refuse or collapse in accuracy, while others remain robust.
The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs
cs.CL 2025-06 conditional novelty 3.0 of 10

A structured survey of LLM safety evaluation that proposes a why/what/where/how taxonomy and catalogs metrics, datasets, benchmarks, evaluators, and frameworks.

Reference graph

Works this paper leans on

213 extracted references · 14 canonical work pages · cited by 2 Pith papers

[23]

Hallucination is inevitable: An innate limitation of large language models,

A. Author and B. Coauthor, “Hallucination is inevitable: An innate limitation of large language models,”Journal of Machine Learning Research, vol. 25, no. 1, pp. 1–30, 2024, theoretical proof that LLM hallucinations arise from fundamental trade-offs in autoregressive generation. [Online]. Available: https://jmlr.org/papers/v25/author2024.html

2024
[1]

Recent advances in natural language processing via large pre-trained language models: A survey,

B. Min, H. Ross, E. Sulem, A. P. B. Veyseh, T. H. Nguyen, O. Sainz, E. Agirre, I. Heintz, and D. Roth, “Recent advances in natural language processing via large pre-trained language models: A survey,”ACM Computing Surveys, vol. 56, no. 2, pp. 1–40, 2023

2023
[3]

Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,

D. Guo, D. Yang, H. Zhang, J. Song, R. Zhang, R. Xu, Q. Zhu, S. Ma, P. Wang, X. Biet al., “Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,”arXiv preprint arXiv:2501.12948, 2025

arXiv 2025
[4]

A comprehensive survey of hallucination mitigation tech- niques in large language models,

S. Tonmoy, S. Zaman, V . Jain, A. Rani, V . Rawte, A. Chadha, and A. Das, “A comprehensive survey of hallucination mitigation tech- niques in large language models,”arXiv preprint arXiv:2401.01313, 2024

arXiv 2024
[5]

hallucination,

“hallucination,” https://www.etymonline.com/word/hallucination, ac- cessed: 2025-03-03

2025
[6]

Des maladies mentales considerees sous les rapports medical, hygi ´enique et medico-legal,

J.-E. D. Esquirol, “Des maladies mentales considerees sous les rapports medical, hygi ´enique et medico-legal,”Bailliere, 1838
[7]

Hallucinations in schizophrenia and parkinson’s disease,

G. F ´enelon, F. Mahieux, R. Huon, and M. Ziegler, “Hallucinations in schizophrenia and parkinson’s disease,”Psychopathology, vol. 33, no. 6, pp. 299–306, 2000. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/11124598/

arXiv 2000
[8]

Six challenges for neural machine transla- tion,

P. Koehn and R. Knowles, “Six challenges for neural machine transla- tion,”arXiv preprint arXiv:1706.03872, 2017

arXiv 2017

Show all 213 references

[9]

On faith- fulness and factuality in abstractive summarization,

J. Maynez, S. Narayan, B. Bohnet, and R. McDonald, “On faith- fulness and factuality in abstractive summarization,”arXiv preprint arXiv:2005.00661, 2020

2005 arXiv
[10]

Survey of hallucination in natural language generation,

Z. Ji, N. Lee, R. Frieske, T. Yu, D. Su, Y . Xu, E. Ishii, Y . J. Bang, A. Madotto, and P. Fung, “Survey of hallucination in natural language generation,”ACM Computing Surveys, vol. 55, no. 12, pp. 1–38, 2023

2023
[11]

Llms will always hallucinate, and we need to live with this,

S. Banerjee, A. Agarwal, and S. Singla, “Llms will always hallucinate, and we need to live with this,”arXiv preprint arXiv:2409.05746, 2024

2024 arXiv
[12]

A survey of hallucination in large foundation models,

V . Rawte, A. Sheth, and A. Das, “A survey of hallucination in large foundation models,”arXiv preprint arXiv:2309.05922, 2023

2023 arXiv
[13]

Siren’s song in the ai ocean: a survey on hal- lucination in large language models,

Y . Zhang, Y . Li, L. Cui, D. Cai, L. Liu, T. Fu, X. Huang, E. Zhao, Y . Zhang, Y . Chenet al., “Siren’s song in the ai ocean: a survey on hal- lucination in large language models,”arXiv preprint arXiv:2309.01219, 2023

2023 arXiv
[14]

A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions,

L. Huang, W. Yu, W. Ma, W. Zhong, Z. Feng, H. Wang, Q. Chen, W. Peng, X. Feng, B. Qinet al., “A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions,”arXiv preprint arXiv:2311.05232, 2023

2023 arXiv
[15]

Hallucination of multimodal large language models: A survey,

Z. Bai, P. Wang, T. Xiao, T. He, Z. Han, Z. Zhang, and M. Z. Shou, “Hallucination of multimodal large language models: A survey,”arXiv preprint arXiv:2404.18930, 2024

2024 arXiv
[16]

A survey on large language model hallucination via a creativity perspective,

X. Jiang, Y . Tian, F. Hua, C. Xu, Y . Wang, and J. Guo, “A survey on large language model hallucination via a creativity perspective,”arXiv preprint arXiv:2402.06647, 2024

2024 arXiv
[17]

The troubling emergence of hallucination in large language models–an extensive definition, quantification, and prescriptive remediations,

V . Rawte, S. Chakraborty, A. Pathak, A. Sarkar, S. Tonmoy, A. Chadha, A. P. Sheth, and A. Das, “The troubling emergence of hallucination in large language models–an extensive definition, quantification, and prescriptive remediations,”arXiv preprint arXiv:2310.04988, 2023

arXiv
[18]

The ai trust paradox: Navigating verisimilitude in advanced language models,

C. Foster-McBride, “The ai trust paradox: Navigating verisimilitude in advanced language models,”Digital Human Assistants, 2024. [Online]. Available: https://www.digitalhumanassistants.com/ai-trust-paradox

2024
[19]

Design, validation, and risk assessment of llm-based generative ai systems operating in the legal sector,

E. Buchicchio, A. De Angelis, A. Moschitta, F. Santoni, L. San Marco, and P. Carbone, “Design, validation, and risk assessment of llm-based generative ai systems operating in the legal sector,” in2024 IEEE International Symposium on Systems Engineering (ISSE). IEEE, 2024, pp. 1–8

2024
[20]

Large language models illuminate a progressive pathway to artificial healthcare assistant: A review,

M. Yuan, P. Bao, J. Yuan, Y . Shen, Z. Chen, Y . Xie, J. Zhao, Y . Chen, L. Zhang, L. Shen, and B. Dong, “Large language models illuminate a progressive pathway to artificial healthcare assistant: A review,”arXiv preprint arXiv:2311.01918, 2023. [Online]. Available: https://ar...

2023 arXiv
[21]

A survey on large language models for critical societal domains: Finance, healthcare, and law,

Z. Z. Chen, J. Ma, X. Zhang, N. Hao, A. Yan, A. Nourbakhsh, X. Yang, J. McAuley, L. Petzold, and W. Y . Wang, “A survey on large language models for critical societal domains: Finance, healthcare, and law,”arXiv preprint arXiv:2405.01769, 2024. [Online]. Available: https://arx...

2024 arXiv
[22]

Ai in education: The risks and opportunities of large language models in learning systems,

X. Chen, Z. Liu, and L. Zhang, “Ai in education: The risks and opportunities of large language models in learning systems,”arXiv preprint arXiv:2404.12218, 2024. [Online]. Available: https://arxiv.org/abs/2404.12218

2024 arXiv
[24]

Hallucinations are inevitable but statistically negligible,

A. Suzuki, Y . He, F. Tian, and Z. Wang, “Hallucinations are inevitable but statistically negligible,”arXiv preprint arXiv:2502.12187, 2025

2025 arXiv
[25]

Calibrated language models must hallucinate,

A. T. Kalai and S. S. Vempala, “Calibrated language models must hallucinate,” inProceedings of the 56th Annual ACM Symposium on Theory of Computing, ser. STOC 2024. New York, NY , USA: Association for Computing Machinery, 2024, p. 160–171. [Online]. Available: https://doi.org/...

2024
[26]

Towards trustworthy llms: a review on debiasing and dehallucinating in large language models,

Z. Lin, S. Guan, W. Zhang, H. Zhang, Y . Li, and H. Zhang, “Towards trustworthy llms: a review on debiasing and dehallucinating in large language models,”Artificial Intelligence Review, vol. 57, no. 9, p. 243, 2024

2024
[27]

Knowledge graphs, large language models, and hallucinations: An nlp perspective,

E. Lavrinovics, R. Biswas, J. Bjerva, and K. Hose, “Knowledge graphs, large language models, and hallucinations: An nlp perspective,”Journal of Web Semantics, vol. 85, p. 100844, 2025

2025
[28]

Med-HALT: Medical domain hallucination test for large language models,

A. Pal, L. K. Umapathi, and M. Sankarasubbu, “Med-HALT: Medical domain hallucination test for large language models,” inProceedings of the 27th Conference on Computational Natural Language Learning (CoNLL), J. Jiang, D. Reitter, and S. Deng, Eds. Singapore: Association for Com...

2023
[29]

Hallucinations in chatgpt: An unreliable tool for learning,

Z. Ahmad, W. Kaiser, and S. Rahim, “Hallucinations in chatgpt: An unreliable tool for learning,”Rupkatha Journal on Interdisciplinary Studies in Humanities, vol. 15, no. 4, p. 12, 2023

2023
[30]

Hallucination mitigation for retrieval- augmented large language models: A review,

W. Zhang and J. Zhang, “Hallucination mitigation for retrieval- augmented large language models: A review,”Mathematics, vol. 13, no. 5, p. 856, 2025

2025
[31]

Halueval: A large-scale hallucination evaluation benchmark for large language models,

J. Li, X. Cheng, W. X. Zhao, J.-Y . Nie, and J.-R. Wen, “Halueval: A large-scale hallucination evaluation benchmark for large language models,”arXiv preprint arXiv:2305.11747, 2023

2023 arXiv
[32]

Defan: Defini- tive answer dataset for llms hallucination evaluation,

A. Rahman, S. Anwar, M. Usman, and A. Mian, “Defan: Defini- tive answer dataset for llms hallucination evaluation,”arXiv preprint arXiv:2406.09155, 2024

2024 arXiv
[33]

Benchmarking hallucination in large language models based on unanswerable math word problem,

Y . Sun, Z. Yin, Q. Guo, J. Wu, X. Qiu, and H. Zhao, “Benchmarking hallucination in large language models based on unanswerable math word problem,”arXiv preprint arXiv:2403.03558, 2024

2024 arXiv
[34]

Fine-grained hal- lucination detection and mitigation in long-form question answering,

R. Sachdeva, Y . Song, M. Iyyer, and I. Gurevych, “Fine-grained hal- lucination detection and mitigation in long-form question answering,” arXiv preprint arXiv:2407.11930, 2024

2024 arXiv
[35]

Truthfulqa: Measuring how models mimic human falsehoods,

S. Lin, J. Hilton, and O. Evans, “Truthfulqa: Measuring how models mimic human falsehoods,”arXiv preprint arXiv:2109.07958, 2021

2021 arXiv
[36]

Unified hallucination detection for multimodal large language models,

X. Chen, C. Wang, Y . Xue, N. Zhang, X. Yang, Q. Li, Y . Shen, J. Gu, and H. Chen, “Unified hallucination detection for multimodal large language models,”arXiv preprint arXiv:2402.03190, 2024

2024 arXiv
[38]

Trans- ferable and efficient non-factual content detection via probe training with offline consistency checking,

X. Zhang, Z. Yao, J. Zhang, K. Yun, J. Yu, J. Li, and J. Tang, “Trans- ferable and efficient non-factual content detection via probe training with offline consistency checking,”arXiv preprint arXiv:2404.06742, 2024

2024 arXiv
[39]

Llm internal states reveal hallucination risk faced with a query,

Z. Ji, D. Chen, E. Ishii, S. Cahyawijaya, Y . Bang, B. Wilie, and P. Fung, “Llm internal states reveal hallucination risk faced with a query,”arXiv preprint arXiv:2407.03282, 2024

2024 arXiv
[40]

Uncertainty estimation and quantification for llms: A simple supervised approach,

L. Liu, Y . Pan, X. Li, and G. Chen, “Uncertainty estimation and quantification for llms: A simple supervised approach,”arXiv preprint arXiv:2404.15993, 2024

2024 arXiv
[41]

Confi- dence under the hood: An investigation into the confidence-probability alignment in large language models,

A. Kumar, R. Morabito, S. Umbet, J. Kabbara, and A. Emami, “Confi- dence under the hood: An investigation into the confidence-probability alignment in large language models,”arXiv preprint arXiv:2405.16282, 2024

2024 arXiv
[42]

Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation,

L. Kuhn, Y . Gal, and S. Farquhar, “Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation,” arXiv preprint arXiv:2302.09664, 2023

2023 arXiv
[43]

Detecting hallucinations in large language models using semantic entropy,

S. Farquhar, J. Kossen, L. Kuhn, and Y . Gal, “Detecting hallucinations in large language models using semantic entropy,”Nature, vol. 630, no. 8017, pp. 625–630, 2024

2024
[44]

On hallucination and predictive uncertainty in conditional language generation,

Y . Xiao and W. Y . Wang, “On hallucination and predictive uncertainty in conditional language generation,”arXiv preprint arXiv:2103.15025, 2021

2021 arXiv
[45]

Inside: Llms’ internal states retain the power of hallucination detection,

C. Chen, K. Liu, Z. Chen, Y . Gu, Y . Wu, M. Tao, Z. Fu, and J. Ye, “Inside: Llms’ internal states retain the power of hallucination detection,”arXiv preprint arXiv:2402.03744, 2024

2024 arXiv
[46]

Knowledge overshadowing causes amalgamated hallucination in large language models,

Y . Zhang, S. Li, J. Liu, P. Yu, Y . R. Fung, J. Li, M. Li, and H. Ji, “Knowledge overshadowing causes amalgamated hallucination in large language models,” 2024. [Online]. Available: https://arxiv.org/abs/2407.08039

2024 arXiv
[47]

Lookback lens: Detecting and mitigating contextual hallucinations in large language models using only attention maps,

Y .-S. Chuang, L. Qiu, C.-Y . Hsieh, R. Krishna, Y . Kim, and J. R. Glass, “Lookback lens: Detecting and mitigating contextual hallucinations in large language models using only attention maps,” inProceedings of the 2024 Conference on Empirical Methods in Natural Language Proc...

2024
[48]

Leveraging graph structures to detect hallucinations in large language models,

N. Nonkes, S. Agaronian, E. Kanoulas, and R. Petcu, “Leveraging graph structures to detect hallucinations in large language models,” inProceedings of TextGraphs-17: Graph-based Methods for Natural Language Processing, D. Ustalov, Y . Gao, A. Panchenko, E. Tutubalina, I. Nikish...

2024
[49]

Uncertainty estimation of large language models in medical question answering,

J. Wu, Y . Yu, and H.-Y . Zhou, “Uncertainty estimation of large language models in medical question answering,”arXiv preprint arXiv:2407.08662, 2024

2024 arXiv
[50]

Rcot: Detecting and rectifying factual inconsistency in reasoning by reversing chain- of-thought,

T. Xue, Z. Wang, Z. Wang, C. Han, P. Yu, and H. Ji, “Rcot: Detecting and rectifying factual inconsistency in reasoning by reversing chain- of-thought,”arXiv preprint arXiv:2305.11499, 2023

2023 arXiv
[51]

Selfcheckgpt: Zero-resource black-box hallucination detection for generative large language mod- els,

P. Manakul, A. Liusie, and M. J. Gales, “Selfcheckgpt: Zero-resource black-box hallucination detection for generative large language mod- els,”arXiv preprint arXiv:2303.08896, 2023

2023 arXiv
[52]

Detecting and mitigating hallucinations in multilingual summarisation,

Y . Qiu, Y . Ziser, A. Korhonen, E. M. Ponti, and S. B. Cohen, “Detecting and mitigating hallucinations in multilingual summarisation,”arXiv preprint arXiv:2305.13632, 2023

2023 arXiv
[53]

Sac3: Reliable hallucination detection in black-box language models via semantic- aware cross-check consistency,

J. Zhang, Z. Li, K. Das, B. A. Malin, and S. Kumar, “Sac3: Reliable hallucination detection in black-box language models via semantic- aware cross-check consistency,”arXiv preprint arXiv:2311.01740, 2023

2023 arXiv
[54]

Mitigating dialogue hallucination for large multi-modal models via adversarial instruction tuning,

D. Park, Z. Qian, G. Han, and S.-N. Lim, “Mitigating dialogue hallucination for large multi-modal models via adversarial instruction tuning,”arXiv preprint arXiv:2403.10492, 2024

2024 arXiv
[55]

Detecting llm hallucination through layer-wise information deficiency: Analysis of unanswerable questions and ambiguous prompts,

H. Kim, A. Bibi, P. Torr, and Y . Gal, “Detecting llm hallucination through layer-wise information deficiency: Analysis of unanswerable questions and ambiguous prompts,” 2024. [Online]. Available: https://arxiv.org/abs/2412.10246

2024
[56]

Teaching models to express their uncertainty in words,

S. Lin, J. Hilton, and O. Evans, “Teaching models to express their uncertainty in words,”arXiv preprint arXiv:2205.14334, 2022

2022 arXiv
[57]

Can llms express their uncertainty? an empirical evaluation of confidence elicitation in llms,

M. Xiong, Z. Hu, X. Lu, Y . Li, J. Fu, J. He, and B. Hooi, “Can llms express their uncertainty? an empirical evaluation of confidence elicitation in llms,”arXiv preprint arXiv:2306.13063, 2023

2023 arXiv
[58]

Can large language models faithfully express their intrinsic uncertainty in words?

G. Yona, R. Aharoni, and M. Geva, “Can large language models faithfully express their intrinsic uncertainty in words?”arXiv preprint arXiv:2405.16908, 2024

2024 arXiv
[59]

Navigating the grey area: Expressions of overconfidence and uncertainty in language models. arxiv abs/2302.13439 (2023),

K. Zhou, D. Jurafsky, and T. Hashimoto, “Navigating the grey area: Expressions of overconfidence and uncertainty in language models. arxiv abs/2302.13439 (2023),” 2023

2023 arXiv
[60]

Hallucana: Fixing llm hallucination with a canary lookahead,

T. Li, E. Dayanik, S. Tyagi, and A. Pierleoni, “Hallucana: Fixing llm hallucination with a canary lookahead,” 2024. [Online]. Available: https://arxiv.org/abs/2412.07965

2024 arXiv
[61]

Calm: Curiosity-driven auditing for large language models,

X. Zheng, L. Wang, Y . Liu, X. Ma, C. Shen, and C. Wang, “Calm: Curiosity-driven auditing for large language models,” 2025. [Online]. Available: https://arxiv.org/abs/2501.02997

2025 arXiv
[62]

Towards reliable mis- information mitigation: Generalization, uncertainty, and gpt-4,

K. Pelrine, A. Imouza, C. Thibault, M. Reksoprodjo, C. Gupta, J. Christoph, J.-F. Godbout, and R. Rabbany, “Towards reliable mis- information mitigation: Generalization, uncertainty, and gpt-4,”arXiv preprint arXiv:2305.14928, 2023. 28

2023 arXiv
[63]

What’s wrong? refining meeting summaries with llm feedback,

F. Kirstein, T. Ruas, and B. Gipp, “What’s wrong? refining meeting summaries with llm feedback,”arXiv preprint arXiv:2407.11919, 2024

2024 arXiv
[64]

Uncovering biases with reflective large language mod- els,

E. Y . Chang, “Uncovering biases with reflective large language mod- els,”arXiv preprint arXiv:2408.13464, 2024

2024 arXiv
[65]

Slm meets llm: Balancing latency, interpretability and con- sistency in hallucination detection,

M. Hu, R. Xu, D. Lei, Y . Li, M. Wang, E. Ching, E. Kamal, and A. Deng, “Slm meets llm: Balancing latency, interpretability and con- sistency in hallucination detection,”arXiv preprint arXiv:2408.12748, 2024

2024 arXiv
[66]

Can llm be a good path planner based on prompt engineering? mitigating the hallucination for path planning,

H. Deng, H. Zhang, J. Ou, and C. Feng, “Can llm be a good path planner based on prompt engineering? mitigating the hallucination for path planning,” 2024. [Online]. Available: https://arxiv.org/abs/2408.13184

2024 arXiv
[67]

Knowledge of knowl- edge: Exploring known-unknowns uncertainty with large language models,

A. Amayuelas, L. Pan, W. Chen, and W. Wang, “Knowledge of knowl- edge: Exploring known-unknowns uncertainty with large language models,”arXiv preprint arXiv:2305.13712, 2023

2023 arXiv
[68]

R-tuning: Instructing large language models to say ‘i don’t know’,

H. Zhang, S. Diao, Y . Lin, Y . Fung, Q. Lian, X. Wang, Y . Chen, H. Ji, and T. Zhang, “R-tuning: Instructing large language models to say ‘i don’t know’,” inProceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Huma...

2024
[69]

Can ai assistants know what they don’t know?

Q. Cheng, T. Sun, X. Liu, W. Zhang, Z. Yin, S. Li, L. Li, K. Chen, and X. Qiu, “Can ai assistants know what they don’t know?”arXiv preprint arXiv:2401.13275, 2024

2024 arXiv
[70]

Linguistic calibration of long-form generations,

N. Band, X. Li, T. Ma, and T. Hashimoto, “Linguistic calibration of long-form generations,” inForty-first International Conference on Machine Learning, 2024

2024
[71]

Sayself: Teaching llms to express confidence with self-reflective rationales,

T. Xu, S. Wu, S. Diao, X. Liu, X. Wang, Y . Chen, and J. Gao, “Sayself: Teaching llms to express confidence with self-reflective rationales,” arXiv preprint arXiv:2405.20974, 2024

2024 arXiv
[72]

Preserving pre-trained features helps cal- ibrate fine-tuned language models,

G. He, J. Chen, and J. Zhu, “Preserving pre-trained features helps cal- ibrate fine-tuned language models,”arXiv preprint arXiv:2305.19249, 2023

2023 arXiv
[73]

Just ask for calibration: Strategies for eliciting calibrated confidence scores from language models fine-tuned with human feedback,

K. Tian, E. Mitchell, A. Zhou, A. Sharma, R. Rafailov, H. Yao, C. Finn, and C. D. Manning, “Just ask for calibration: Strategies for eliciting calibrated confidence scores from language models fine-tuned with human feedback,”arXiv preprint arXiv:2305.14975, 2023

2023 arXiv
[74]

Few-shot recalibration of language models,

X. L. Li, U. Khandelwal, and K. Guu, “Few-shot recalibration of language models,”arXiv preprint arXiv:2403.18286, 2024

2024 arXiv
[75]

Investigating the factual knowledge boundary of large language models with retrieval augmentation,

R. Ren, Y . Wang, Y . Qu, W. X. Zhao, J. Liu, H. Tian, H. Wu, J.- R. Wen, and H. Wang, “Investigating the factual knowledge boundary of large language models with retrieval augmentation,”arXiv preprint arXiv:2307.11019, 2023

2023 arXiv
[76]

Do large language models know what they don’t know?

Z. Yin, Q. Sun, Q. Guo, J. Wu, X. Qiu, and X. Huang, “Do large language models know what they don’t know?”arXiv preprint arXiv:2305.18153, 2023

2023 arXiv
[77]

Halo: Hallucination analysis and learning optimization to empower llms with retrieval-augmented context for guided clinical decision making,

S. Anjum, H. Zhang, W. Zhou, E. J. Paek, X. Zhao, and Y . Feng, “Halo: Hallucination analysis and learning optimization to empower llms with retrieval-augmented context for guided clinical decision making,”arXiv preprint arXiv:2409.10011, 2024

2024 arXiv
[78]

An information bottleneck perspective for effec- tive noise filtering on retrieval-augmented generation,

K. Zhu, X. Feng, X. Du, Y . Gu, W. Yu, H. Wang, Q. Chen, Z. Chu, J. Chen, and B. Qin, “An information bottleneck perspective for effec- tive noise filtering on retrieval-augmented generation,”arXiv preprint arXiv:2406.01549, 2024

2024
[79]

Un- supervised information refinement training of large language models for retrieval-augmented generation,

S. Xu, L. Pang, M. Yu, F. Meng, H. Shen, X. Cheng, and J. Zhou, “Un- supervised information refinement training of large language models for retrieval-augmented generation,”arXiv preprint arXiv:2402.18150, 2024

2024 arXiv
[80]

Grounding lan- guage model with chunking-free in-context retrieval,

H. Qian, Z. Liu, K. Mao, Y . Zhou, and Z. Dou, “Grounding lan- guage model with chunking-free in-context retrieval,”arXiv preprint arXiv:2402.09760, 2024

2024 arXiv
[81]

Redeep: Detecting hallucination in retrieval-augmented generation via mechanistic interpretability,

Z. Sun, X. Zang, K. Zheng, Y . Song, J. Xu, X. Zhang, W. Yu, and H. Li, “Redeep: Detecting hallucination in retrieval-augmented generation via mechanistic interpretability,”arXiv preprint arXiv:2410.11414, 2024

2024
[82]

Mitigating large language model hallucinations via autonomous knowledge graph- based retrofitting,

X. Guan, Y . Liu, H. Lin, Y . Lu, B. He, X. Han, and L. Sun, “Mitigating large language model hallucinations via autonomous knowledge graph- based retrofitting,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 16, 2024, pp. 18 126–18 134

2024
[83]

Unikgqa: Unified retrieval and reasoning for solving multi-hop question answering over knowledge graph,

J. Jiang, K. Zhou, W. X. Zhao, and J.-R. Wen, “Unikgqa: Unified retrieval and reasoning for solving multi-hop question answering over knowledge graph,”arXiv preprint arXiv:2212.00959, 2022

2022 arXiv
[84]

Knowledge graph error detection with con- trastive confidence adaption,

X. Liu, Y . Liu, and W. Hu, “Knowledge graph error detection with con- trastive confidence adaption,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 8, 2024, pp. 8824–8831

2024
[85]

Two-stage generative question answering on temporal knowledge graph using large language models,

Y . Gao, L. Qiao, Z. Kan, Z. Wen, Y . He, and D. Li, “Two-stage generative question answering on temporal knowledge graph using large language models,”arXiv preprint arXiv:2402.16568, 2024

2024 arXiv
[86]

Combining llms and knowledge graphs to reduce hallucinations in question answering,

L. Pusch and T. O. Conrad, “Combining llms and knowledge graphs to reduce hallucinations in question answering,”arXiv preprint arXiv:2409.04181, 2024

2024
[87]

Chain-of-thought prompting elicits reasoning in large language models,

J. Wei, X. Wang, D. Schuurmans, M. Bosma, F. Xia, E. Chi, Q. V . Le, D. Zhouet al., “Chain-of-thought prompting elicits reasoning in large language models,”Advances in neural information processing systems, vol. 35, pp. 24 824–24 837, 2022

2022
[88]

Least-to-most prompting enables complex reasoning in large language models,

D. Zhou, N. Sch ¨arli, L. Hou, J. Wei, N. Scales, X. Wang, D. Schu- urmans, C. Cui, O. Bousquet, Q. Leet al., “Least-to-most prompting enables complex reasoning in large language models,”arXiv preprint arXiv:2205.10625, 2022

2022 arXiv
[89]

Automatic chain of thought prompting in large language models,

Z. Zhang, A. Zhang, M. Li, and A. Smola, “Automatic chain of thought prompting in large language models,”arXiv preprint arXiv:2210.03493, 2022

2022 arXiv
[90]

Measuring and narrowing the compositionality gap in language mod- els,

O. Press, M. Zhang, S. Min, L. Schmidt, N. A. Smith, and M. Lewis, “Measuring and narrowing the compositionality gap in language mod- els,”arXiv preprint arXiv:2210.03350, 2022

2022 arXiv
[91]

Self-contradictory hallucinations of large language models: Evaluation, detection and mitigation,

N. M ¨undler, J. He, S. Jenko, and M. Vechev, “Self-contradictory hallucinations of large language models: Evaluation, detection and mitigation,”arXiv preprint arXiv:2305.15852, 2023

2023 arXiv
[92]

Iao prompting: Making knowledge flow explicit in llms through structured reasoning templates,

A. Diallo, A. Bikakis, L. Dickens, A. Hunter, and R. Miller, “Iao prompting: Making knowledge flow explicit in llms through structured reasoning templates,” 2025. [Online]. Available: https://arxiv.org/abs/2502.03080

2025 arXiv
[93]

Self-alignment for factuality: Mitigating hallucinations in llms via self-evaluation,

X. Zhang, B. Peng, Y . Tian, J. Zhou, L. Jin, L. Song, H. Mi, and H. Meng, “Self-alignment for factuality: Mitigating hallucinations in llms via self-evaluation,”arXiv preprint arXiv:2402.09267, 2024

2024 arXiv
[94]

Self-contrast: Better reflection through inconsistent solving perspectives,

W. Zhang, Y . Shen, L. Wu, Q. Peng, J. Wang, Y . Zhuang, and W. Lu, “Self-contrast: Better reflection through inconsistent solving perspectives,” 2024. [Online]. Available: https://arxiv.org/abs/2401.02009

2024 arXiv
[95]

Adaptation with self-evaluation to improve selective prediction in llms,

J. Chen, J. Yoon, S. Ebrahimi, S. O. Arik, T. Pfister, and S. Jha, “Adaptation with self-evaluation to improve selective prediction in llms,”arXiv preprint arXiv:2310.11689, 2023

2023 arXiv
[96]

Queryagent: A reliable and efficient reasoning framework with environmental feedback based self-correction,

X. Huang, S. Cheng, S. Huang, J. Shen, Y . Xu, C. Zhang, and Y . Qu, “Queryagent: A reliable and efficient reasoning framework with environmental feedback based self-correction,”arXiv preprint arXiv:2403.11886, 2024

2024 arXiv
[97]

Is there no such thing as a bad question? h4r: Hallucibot for ratiocination, rewriting, ranking, and routing,

W. Watson, N. Cho, and N. Srishankar, “Is there no such thing as a bad question? h4r: Hallucibot for ratiocination, rewriting, ranking, and routing,” 2024. [Online]. Available: https://arxiv.org/abs/2404.12535

2024 arXiv
[98]

Lora: Low-rank adaptation of large language models

E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, W. Chenet al., “Lora: Low-rank adaptation of large language models.” ICLR, vol. 1, no. 2, p. 3, 2022

2022
[99]

Knowledgeable parameter efficient tuning network for commonsense question answer- ing,

Z. Zhao, L. Hu, H. Zhao, Y . Shao, and Y . Wang, “Knowledgeable parameter efficient tuning network for commonsense question answer- ing,” inProceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2023, pp. 9051– 9063

2023
[100]

From yes-men to truth-tellers: Addressing sycophancy in large language models with pinpoint tuning,

W. Chen, Z. Huang, L. Xie, B. Lin, H. Li, L. Lu, X. Tian, D. Cai, Y . Zhang, W. Wanet al., “From yes-men to truth-tellers: Addressing sycophancy in large language models with pinpoint tuning,”arXiv preprint arXiv:2409.01658, 2024

2024 arXiv
[101]

Uncer- tainty aware learning for language model alignment,

Y . Wang, R. Zheng, L. Ding, Q. Zhang, D. Lin, and D. Tao, “Uncer- tainty aware learning for language model alignment,”arXiv preprint arXiv:2406.04854, 2024

2024 arXiv
[102]

Alphaedit: Null-space constrained knowledge editing for language models,

J. Fang, H. Jiang, K. Wang, Y . Ma, X. Wang, X. He, and T.-s. Chua, “Alphaedit: Null-space constrained knowledge editing for language models,”arXiv preprint arXiv:2410.02355, 2024

2024 arXiv
[103]

Llms know more than they show: On the intrinsic representation of llm hallucinations,

H. Orgad, M. Toker, Z. Gekhman, R. Reichart, I. Szpektor, H. Kotek, and Y . Belinkov, “Llms know more than they show: On the intrinsic representation of llm hallucinations,”arXiv preprint arXiv:2410.02707, 2024

2024 arXiv
[104]

Interpreting arithmetic mechanism in large language models through comparative neuron analysis,

Z. Yu and S. Ananiadou, “Interpreting arithmetic mechanism in large language models through comparative neuron analysis,”arXiv preprint arXiv:2409.14144, 2024

2024 arXiv
[105]

Selfie: Self-interpretation of large language model embeddings,

H. Chen, C. V ondrick, and C. Mao, “Selfie: Self-interpretation of large language model embeddings,”arXiv preprint arXiv:2403.10949, 2024

2024 arXiv
[106]

Adaptive large language models by layerwise attention shortcuts,

P. Verma and M. Pilanci, “Adaptive large language models by layerwise attention shortcuts,”arXiv preprint arXiv:2409.10870, 2024

2024 arXiv
[107]

Inference- time intervention: Eliciting truthful answers from a language model,

K. Li, O. Patel, F. Vi ´egas, H. Pfister, and M. Wattenberg, “Inference- time intervention: Eliciting truthful answers from a language model,” Advances in Neural Information Processing Systems, vol. 36, 2024

2024
[108]

Truthx: Alleviating hallucinations by editing large language models in truthful space,

S. Zhang, T. Yu, and Y . Feng, “Truthx: Alleviating hallucinations by editing large language models in truthful space,”arXiv preprint arXiv:2402.17811, 2024. 29

2024 arXiv
[109]

Internal consistency and self-feedback in large language models: A survey,

X. Liang, S. Song, Z. Zheng, H. Wang, Q. Yu, X. Li, R.-H. Li, F. Xiong, and Z. Li, “Internal consistency and self-feedback in large language models: A survey,”arXiv preprint arXiv:2407.14507, 2024

2024 arXiv
[110]

Understanding intermediate layers using linear classifier probes,

G. Alain, “Understanding intermediate layers using linear classifier probes,”arXiv preprint arXiv:1610.01644, 2016

2016 arXiv
[111]

Interpretability in the wild: a circuit for indirect object identification in gpt-2 small,

K. Wang, A. Variengien, A. Conmy, B. Shlegeris, and J. Steinhardt, “Interpretability in the wild: a circuit for indirect object identification in gpt-2 small,”arXiv preprint arXiv:2211.00593, 2022

2022 arXiv
[112]

Trusting your evidence: Hallucinate less with context-aware decoding,

W. Shi, X. Han, M. Lewis, Y . Tsvetkov, L. Zettlemoyer, and S. W.- t. Yih, “Trusting your evidence: Hallucinate less with context-aware decoding,”arXiv preprint arXiv:2305.14739, 2023

2023 arXiv
[113]

Lookback lens: Detecting and mitigating contextual hallucinations in large language models using only attention maps,

Y .-S. Chuang, L. Qiu, C.-Y . Hsieh, R. Krishna, Y . Kim, and J. Glass, “Lookback lens: Detecting and mitigating contextual hallucinations in large language models using only attention maps,”arXiv preprint arXiv:2407.07071, 2024

2024 arXiv
[114]

Paying more attention to source context: Mitigating unfaithful translations from large language model,

H. Zhang, K. Chen, X. Bai, Y . Xiang, and M. Zhang, “Paying more attention to source context: Mitigating unfaithful translations from large language model,”arXiv preprint arXiv:2406.07036, 2024

2024 arXiv
[115]

Enhancing contextual understanding in large language models through contrastive decoding,

Z. Zhao, E. Monti, J. Lehmann, and H. Assem, “Enhancing contextual understanding in large language models through contrastive decoding,” arXiv preprint arXiv:2405.02750, 2024

2024 arXiv
[116]

Decore: decoding by contrasting retrieval heads to mitigate hallucinations,

A. P. Gema, C. Jin, A. Abdulaal, T. Diethe, P. Teare, B. Alex, P. Min- ervini, and A. Saseendran, “Decore: decoding by contrasting retrieval heads to mitigate hallucinations,”arXiv preprint arXiv:2410.18860, 2024

2024 arXiv
[117]

Uscd: Improving code generation of llms by uncertainty-aware selective contrastive decoding,

S. Wang, L. Ding, L. Shen, Y . Luo, Z. He, W. Yu, and D. Tao, “Uscd: Improving code generation of llms by uncertainty-aware selective contrastive decoding,”arXiv preprint arXiv:2409.05923, 2024

2024 arXiv
[118]

Alleviating hallucinations of large language models through induced hallucinations,

Y . Zhang, L. Cui, W. Bi, and S. Shi, “Alleviating hallucinations of large language models through induced hallucinations,”arXiv preprint arXiv:2312.15710, 2023

2023 arXiv
[119]

Dola: Decoding by contrasting layers improves factuality in large language models,

Y .-S. Chuang, Y . Xie, H. Luo, Y . Kim, J. Glass, and P. He, “Dola: Decoding by contrasting layers improves factuality in large language models,”arXiv preprint arXiv:2309.03883, 2023

2023 arXiv
[120]

Quantifying and attributing the hallucination of large language models via association analysis,

L. Du, Y . Wang, X. Xing, Y . Ya, X. Li, X. Jiang, and X. Fang, “Quantifying and attributing the hallucination of large language models via association analysis,”arXiv preprint arXiv:2309.05217, 2023

2023 arXiv
[121]

Memory augmented large language models are com- putationally universal,

D. Schuurmans, “Memory augmented large language models are com- putationally universal,”arXiv preprint arXiv:2301.04589, 2023

2023 arXiv
[122]

Go simple and pre-train on domain-specific corpora: On the role of training data for text classification,

A. Edwards, J. Camacho-Collados, H. De Ribaupierre, and A. Preece, “Go simple and pre-train on domain-specific corpora: On the role of training data for text classification,” inProceedings of the 28th international conference on computational linguistics, 2020, pp. 5522– 5529

2020
[123]

Flame: Factuality-aware alignment for large language models,

S.-C. Lin, L. Gao, B. Oguz, W. Xiong, J. Lin, W.-t. Yih, and X. Chen, “Flame: Factuality-aware alignment for large language models,”arXiv preprint arXiv:2405.01525, 2024

2024 arXiv
[124]

Enhancing reinforcement learning with label-sensitive reward for natural language understanding,

K. Liao, S. Li, M. Zhao, L. Liu, M. Xue, Z. Hu, H. Han, and C. Yin, “Enhancing reinforcement learning with label-sensitive reward for natural language understanding,”arXiv preprint arXiv:2405.19763, 2024

2024 arXiv
[125]

A systematic comparison of syllogistic reasoning in humans and language models,

T. Eisape, M. H. Tessler, I. Dasgupta, F. Sha, S. van Steenkiste, and T. Linzen, “A systematic comparison of syllogistic reasoning in humans and language models,”arXiv preprint arXiv:2311.00445, 2023

2023 arXiv
[126]

Gradual syntactic label replacement for language model pre-training,

Y . Wang, Y . Zhang, P. Li, and Y . Liu, “Gradual syntactic label replacement for language model pre-training,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 32, pp. 486–496, 2023

2023
[127]

Unsupervised im- provement of factual knowledge in language models,

N. Sadeq, B. Kang, P. Lamba, and J. McAuley, “Unsupervised im- provement of factual knowledge in language models,”arXiv preprint arXiv:2304.01597, 2023

2023 arXiv
[128]

Pre-training text representations as meta learning,

S. Lv, Y . Wang, D. Guo, D. Tang, N. Duan, F. Zhu, M. Gong, L. Shou, R. Ma, D. Jianget al., “Pre-training text representations as meta learning,”arXiv preprint arXiv:2004.05568, 2020

2004 arXiv
[129]

Getting more juice out of the sft data: Reward learning from human demonstration improves sft for llm alignment,

J. Li, S. Zeng, H.-T. Wai, C. Li, A. Garcia, and M. Hong, “Getting more juice out of the sft data: Reward learning from human demonstration improves sft for llm alignment,”arXiv preprint arXiv:2405.17888, 2024

2024 arXiv
[130]

The wisdom of hindsight makes language models better instruction followers,

T. Zhang, F. Liu, J. Wong, P. Abbeel, and J. E. Gonzalez, “The wisdom of hindsight makes language models better instruction followers,” in International Conference on Machine Learning. PMLR, 2023, pp. 41 414–41 428

2023
[131]

Sample efficient reinforcement learning from human feedback via active exploration,

V . Mehta, V . Das, O. Neopane, Y . Dai, I. Bogunovic, J. Schneider, and W. Neiswanger, “Sample efficient reinforcement learning from human feedback via active exploration,” 2023. [Online]. Available: https://arxiv.org/abs/2312.00267

2023 arXiv
[132]

Fine-tuning llms for multi- turn dialogues: Optimizing cross-entropy loss with kl divergence for all rounds of responses,

Z. Teng, Y . Song, X. Ye, and Y . Ouyang, “Fine-tuning llms for multi- turn dialogues: Optimizing cross-entropy loss with kl divergence for all rounds of responses,” inProceedings of the 2024 16th International Conference on Machine Learning and Computing, 2024, pp. 128–133

2024
[133]

An empirical study of catastrophic forgetting in large language models during continual fine-tuning,

Y . Luo, Z. Yang, F. Meng, Y . Li, J. Zhou, and Y . Zhang, “An empirical study of catastrophic forgetting in large language models during continual fine-tuning,”arXiv preprint arXiv:2308.08747, 2023

2023 arXiv
[134]

Training language models to follow instructions with human feedback,

L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. Wainwright, P. Mishkin, C. Zhang, S. Agarwal, K. Slama, A. Rayet al., “Training language models to follow instructions with human feedback,”Advances in neural information processing systems, vol. 35, pp. 27 730–27 744, 2022

2022
[135]

Learning to summarize from human feedback,

F. Liuet al., “Learning to summarize from human feedback,” in Proceedings of the 58th Annual Meeting of the Association for Com- putational Linguistics, 2020

2020
[136]

Superhf: Supervised iterative learning from human feedback,

G. Mukobi, P. Chatain, S. Fong, R. Windesheim, G. Kutyniok, K. Bha- tia, and S. Alberti, “Superhf: Supervised iterative learning from human feedback,”arXiv preprint arXiv:2310.16763, 2023

2023 arXiv
[137]

Secrets of rlhf in large language models part i: Ppo,

R. Zheng, S. Dou, S. Gao, Y . Hua, W. Shen, B. Wang, Y . Liu, S. Jin, Q. Liu, Y . Zhouet al., “Secrets of rlhf in large language models part i: Ppo,”arXiv preprint arXiv:2307.04964, 2023

2023 arXiv
[138]

Reinforcement learning for ques- tion answering in programming domain using public community scor- ing as a human feedback,

A. Gorbatovski and S. Kovalchuk, “Reinforcement learning for ques- tion answering in programming domain using public community scor- ing as a human feedback,”arXiv preprint arXiv:2401.10882, 2024

2024 arXiv
[139]

Teach- ing large language models to reason with reinforcement learning,

A. Havrilla, Y . Du, S. C. Raparthy, C. Nalmpantis, J. Dwivedi-Yu, M. Zhuravinskyi, E. Hambro, S. Sukhbaatar, and R. Raileanu, “Teach- ing large language models to reason with reinforcement learning,” arXiv preprint arXiv:2403.04642, 2024

2024 arXiv
[140]

A baseline analysis of reward models’ ability to accurately analyze foundation models under distribution shift,

B. Pikus, W. LeVine, T. Chen, and S. Hendryx, “A baseline analysis of reward models’ ability to accurately analyze foundation models under distribution shift,”arXiv preprint arXiv:2311.14743, 2023

2023 arXiv
[141]

Lacie: Listener-aware finetuning for confidence calibration in large language models,

E. Stengel-Eskin, P. Hase, and M. Bansal, “Lacie: Listener-aware finetuning for confidence calibration in large language models,”arXiv preprint arXiv:2405.21028, 2024

2024 arXiv
[142]

Why LLMs hallucinate, and how to get (evidential) closure: Perceptual, intensional, and extensional learning for faithful natural language generation,

A. Bouyamourn, “Why LLMs hallucinate, and how to get (evidential) closure: Perceptual, intensional, and extensional learning for faithful natural language generation,” inProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, H. Bouamor, J. Pino...

2023
[143]

Hallucination is inevitable: An innate limitation of large language models,

Z. Xu, S. Jain, and M. Kankanhalli, “Hallucination is inevitable: An innate limitation of large language models,”arXiv preprint arXiv:2401.11817, 2024

2024 arXiv
[144]

G ¨odel’s incompleteness theorems,

P. Raatikainen, “G ¨odel’s incompleteness theorems,”Stanford encyclo- pedia of philosophy, no. Winter 2013, 2013

2013
[145]

Rogers Jr,Theory of recursive functions and effective computability

H. Rogers Jr,Theory of recursive functions and effective computability. MIT press, 1987

1987
[146]

Scaling laws for neural language models,

J. Kaplan, S. McCandlish, T. Henighan, T. B. Brown, B. Chess, R. Child, S. Gray, A. Radford, J. Wu, and D. Amodei, “Scaling laws for neural language models,”arXiv preprint arXiv:2001.08361, 2020

2001 arXiv
[147]

Does fine-tuning llms on new knowledge encourage hallucinations?

Z. Gekhman, G. Yona, R. Aharoni, M. Eyal, A. Feder, R. Reichart, and J. Herzig, “Does fine-tuning llms on new knowledge encourage hallucinations?”arXiv preprint arXiv:2405.05904, 2024

2024 arXiv
[148]

Schrodinger’s memory: Large language models,

W. Wang and Q. Li, “Schrodinger’s memory: Large language models,” arXiv preprint arXiv:2409.10482, 2024

2024 arXiv
[149]

When llms meet cunning questions: A fallacy understanding bench- mark for large language models,

Y . Li, Q. Zhou, Y . Luo, S. Ma, Y . Li, H.-T. Zheng, X. Hu, and P. S. Yu, “When llms meet cunning questions: A fallacy understanding bench- mark for large language models,”arXiv preprint arXiv:2402.11100, 2024

2024 arXiv
[150]

Coig-cqia: Quality is all you need for chinese instruction fine-tuning,

Y . Bai, X. Du, Y . Liang, Y . Jin, Z. Liu, J. Zhou, T. Zheng, X. Zhang, N. Ma, Z. Wanget al., “Coig-cqia: Quality is all you need for chinese instruction fine-tuning,”arXiv preprint arXiv:2403.18058, 2024

2024 arXiv
[151]

Re-examining calibration: The case of question answering,

C. Si, C. Zhao, S. Min, and J. Boyd-Graber, “Re-examining calibration: The case of question answering,”arXiv preprint arXiv:2205.12507, 2022

2022 arXiv
[152]

Uncertainty in language models: Assessment through rank-calibration,

X. Huang, S. Li, M. Yu, M. Sesia, H. Hassani, I. Lee, O. Bastani, and E. Dobriban, “Uncertainty in language models: Assessment through rank-calibration,”arXiv preprint arXiv:2404.03163, 2024

2024 arXiv
[153]

Calibrating the confidence of large language models by eliciting fidelity,

M. Zhang, M. Huang, R. Shi, L. Guo, C. Peng, P. Yan, Y . Zhou, and X. Qiu, “Calibrating the confidence of large language models by eliciting fidelity,”arXiv preprint arXiv:2404.02655, 2024

2024 arXiv
[154]

Uncertainty-based abstention in llms improves safety and reduces hallucinations,

C. Tomani, K. Chaudhuri, I. Evtimov, D. Cremers, and M. Ibrahim, “Uncertainty-based abstention in llms improves safety and reduces hallucinations,” 2024. [Online]. Available: https://arxiv.org/abs/2404.10960 30

2024 arXiv
[157]

Halo: Estimation and reduction of halluci- nations in open-source weak large language models,

M. Elaraby, M. Lu, J. Dunn, X. Zhang, Y . Wang, S. Liu, P. Tian, Y . Wang, and Y . Wang, “Halo: Estimation and reduction of halluci- nations in open-source weak large language models,”arXiv preprint arXiv:2308.11764, 2023

2023 arXiv
[158]

The internal state of an llm knows when it’s lying,

A. Azaria and T. Mitchell, “The internal state of an llm knows when it’s lying,”arXiv preprint arXiv:2304.13734, 2023

2023 arXiv
[159]

Discursive socratic questioning: Evaluating the faithfulness of language models’ under- standing of discourse relations,

Y . Miao, H. Liu, W. Lei, N. Chen, and M.-Y . Kan, “Discursive socratic questioning: Evaluating the faithfulness of language models’ under- standing of discourse relations,” inProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Lo...

2024
[160]

Faithdial: A faithful benchmark for information- seeking dialogue,

N. Dziri, E. Kamalloo, S. Milton, O. Zaiane, M. Yu, E. M. Ponti, and S. Reddy, “Faithdial: A faithful benchmark for information- seeking dialogue,”Transactions of the Association for Computational Linguistics, vol. 10, pp. 1473–1490, 2022

2022
[161]

Instruction-following evaluation for large language models,

J. Zhou, T. Lu, S. Mishra, S. Brahma, S. Basu, Y . Luan, D. Zhou, and L. Hou, “Instruction-following evaluation for large language models,” arXiv preprint arXiv:2311.07911, 2023

2023 arXiv
[162]

Did aristotle use a laptop? a question answering benchmark with implicit reasoning strategies,

M. Geva, D. Khashabi, E. Segal, T. Khot, D. Roth, and J. Berant, “Did aristotle use a laptop? a question answering benchmark with implicit reasoning strategies,”Transactions of the Association for Computational Linguistics, vol. 9, pp. 346–361, 2021

2021
[163]

Triviaqa: A large scale distantly supervised challenge dataset for reading comprehen- sion,

M. Joshi, E. Choi, D. S. Weld, and L. Zettlemoyer, “Triviaqa: A large scale distantly supervised challenge dataset for reading comprehen- sion,”arXiv preprint arXiv:1705.03551, 2017

2017 arXiv
[164]

Hotpotqa: A dataset for diverse, explain- able multi-hop question answering,

Z. Yang, P. Qi, S. Zhang, Y . Bengio, W. W. Cohen, R. Salakhutdi- nov, and C. D. Manning, “Hotpotqa: A dataset for diverse, explain- able multi-hop question answering,”arXiv preprint arXiv:1809.09600, 2018

2018 arXiv
[165]

The hallurag dataset: Detecting closed- domain hallucinations in rag applications using an llm’s internal states,

F. Ridder and M. Schilling, “The hallurag dataset: Detecting closed- domain hallucinations in rag applications using an llm’s internal states,” 2024. [Online]. Available: https://arxiv.org/abs/2412.17056

2024 arXiv
[166]

Hallucination detection: Robustly discerning reliable answers in large language models,

Y . Chen, Q. Fu, Y . Yuan, Z. Wen, G. Fan, D. Liu, D. Zhang, Z. Li, and Y . Xiao, “Hallucination detection: Robustly discerning reliable answers in large language models,” 2024. [Online]. Available: https://arxiv.org/abs/2407.04121

2024 arXiv
[167]

Anah-v2: Scaling analytical hallucination annotation of large language models,

Y . Gu, Z. Ji, W. Zhang, C. Lyu, D. Lin, and K. Chen, “Anah-v2: Scaling analytical hallucination annotation of large language models,”
[168]

Training on the benchmark is not all you need,

S. Ni, X. Kong, C. Li, X. Hu, R. Xu, J. Zhu, and M. Yang, “Training on the benchmark is not all you need,”arXiv preprint arXiv:2409.01790, 2024

2024 arXiv
[169]

Cost-effective hallucination detection for llms,

S. Valentin, J. Fu, G. Detommaso, S. Xu, G. Zappella, and B. Wang, “Cost-effective hallucination detection for llms,” 2024. [Online]. Available: https://arxiv.org/abs/2407.21424

2024 arXiv
[170]

Discovering latent knowledge in language models without supervision,

C. Burns, H. Ye, D. Klein, and J. Steinhardt, “Discovering latent knowledge in language models without supervision,”arXiv preprint arXiv:2212.03827, 2022

2022 arXiv
[171]

” i’m not sure, but

S. S. Kim, Q. V . Liao, M. V orvoreanu, S. Ballard, and J. W. Vaughan, “” i’m not sure, but...”: Examining the impact of large language models’ uncertainty expression on user reliance and trust,” inThe 2024 ACM Conference on Fairness, Accountability, and Transparency, 2024, pp...

2024
[172]

How can we know when language models know? on the calibration of language models for ques- tion answering,

Z. Jiang, J. Araki, H. Ding, and G. Neubig, “How can we know when language models know? on the calibration of language models for ques- tion answering,”Transactions of the Association for Computational Linguistics, vol. 9, pp. 962–977, 2021

2021
[173]

Alleviating hallucinations in large language models with scepticism modeling,

Y . Wu, Y . Wang, T. Chen, C. Liu, N. Xi, Q. Gu, H. Lei, Z. Jiang, Y . Chen, and L. Ji, “Alleviating hallucinations in large language models with scepticism modeling,”arXiv preprint arXiv:2409.06601, 2024

2024 arXiv
[174]

Improving the reliability of large language models by leveraging uncertainty-aware in-context learning,

Y . Yang, H. Li, Y . Wang, and Y . Wang, “Improving the reliability of large language models by leveraging uncertainty-aware in-context learning,”arXiv preprint arXiv:2310.04782, 2023

2023 arXiv
[175]

Chain-of-knowledge: Grounding large language models via dynamic knowledge adapting over heterogeneous sources,

X. Li, R. Zhao, Y . K. Chia, B. Ding, S. Joty, S. Poria, and L. Bing, “Chain-of-knowledge: Grounding large language models via dynamic knowledge adapting over heterogeneous sources,”arXiv preprint arXiv:2305.13269, 2023

2023 arXiv
[176]

M-rag: Reinforcing large language model performance through retrieval-augmented gener- ation with multiple partitions,

Z. Wang, S. X. Teo, J. Ouyang, Y . Xu, and W. Shi, “M-rag: Reinforcing large language model performance through retrieval-augmented gener- ation with multiple partitions,”arXiv preprint arXiv:2405.16420, 2024

2024 arXiv
[177]

On the role of long-tail knowledge in retrieval augmented large language models,

D. Li, J. Yan, T. Zhang, C. Wang, X. He, L. Huang, H. Xue, and J. Huang, “On the role of long-tail knowledge in retrieval augmented large language models,”arXiv preprint arXiv:2406.16367, 2024

2024 arXiv
[178]

Knowledge card: Filling llms’ knowledge gaps with plug-in special- ized language models,

S. Feng, W. Shi, Y . Bai, V . Balachandran, T. He, and Y . Tsvetkov, “Knowledge card: Filling llms’ knowledge gaps with plug-in special- ized language models,”arXiv preprint arXiv:2305.09955, 2023

2023 arXiv
[179]

Listt5: Listwise reranking with fusion-in-decoder improves zero-shot retrieval,

S. Yoon, E. Lee, J. Kim, Y . Kim, H. Yun, and S.-w. Hwang, “Listt5: Listwise reranking with fusion-in-decoder improves zero-shot retrieval,”arXiv preprint arXiv:2402.15838, 2024

2024 arXiv
[180]

Ragtruth: A hallucination corpus for developing trustworthy retrieval-augmented language models,

C. Niu, Y . Wu, J. Zhu, S. Xu, K. Shum, R. Zhong, J. Song, and T. Zhang, “Ragtruth: A hallucination corpus for developing trustworthy retrieval-augmented language models,”arXiv preprint arXiv:2401.00396, 2023

2023 arXiv
[181]

Hallucination-free? assessing the reliability of leading ai legal research tools; 2024,

V . Magesh, F. Surani, M. Dahl, M. Suzgun, C. D. Manning, and D. E. Ho, “Hallucination-free? assessing the reliability of leading ai legal research tools; 2024,”URL https://arxiv. org/abs/2405.20362, 2024

2024 arXiv
[182]

Knowledge graphs,

A. Hogan, E. Blomqvist, M. Cochez, C. d’Amato, G. D. Melo, C. Gutierrez, S. Kirrane, J. E. L. Gayo, R. Navigli, S. Neumaieret al., “Knowledge graphs,”ACM Computing Surveys (Csur), vol. 54, no. 4, pp. 1–37, 2021

2021
[183]

Can knowl- edge graphs reduce hallucinations in llms?: A survey,

G. Agrawal, T. Kumarage, Z. Alghamdi, and H. Liu, “Can knowl- edge graphs reduce hallucinations in llms?: A survey,”arXiv preprint arXiv:2311.07914, 2023

2023 arXiv
[184]

Holmes: Hyper-relational knowledge graphs for multi-hop question answering using llms,

P. Panda, A. Agarwal, C. Devaguptapu, M. Kaulet al., “Holmes: Hyper-relational knowledge graphs for multi-hop question answering using llms,”arXiv preprint arXiv:2406.06027, 2024

2024 arXiv
[185]

Knowformer: Revisit- ing transformers for knowledge graph reasoning,

J. Liu, Q. Mao, W. Jiang, and J. Li, “Knowformer: Revisit- ing transformers for knowledge graph reasoning,”arXiv preprint arXiv:2409.12865, 2024

2024 arXiv
[186]

Kg-rag: Bridging the gap between knowledge and creativity,

D. Sanmartin, “Kg-rag: Bridging the gap between knowledge and creativity,”arXiv preprint arXiv:2405.12035, 2024

2024 arXiv
[187]

Grapheval: A knowledge-graph based llm hallucination evaluation framework,

H. Sansford, N. Richardson, H. P. Maretic, and J. N. Saada, “Grapheval: A knowledge-graph based llm hallucination evaluation framework,” arXiv preprint arXiv:2407.10793, 2024

2024 arXiv
[188]

Memory-based model editing at scale,

E. Mitchell, C. Lin, A. Bosselut, C. D. Manning, and C. Finn, “Memory-based model editing at scale,” inICML, vol. 162. PMLR, 2022, pp. 15 817–15 831

2022
[189]

Prefix-tuning: Optimizing continuous prompts for generation,

X. L. Li and P. Liang, “Prefix-tuning: Optimizing continuous prompts for generation,”arXiv preprint arXiv:2101.00190, 2021

2021 arXiv
[190]

Teaching language models to hallucinate less with synthetic tasks,

E. Jones, H. Palangi, C. Sim ˜oes, V . Chandrasekaran, S. Mukherjee, A. Mitra, A. Awadallah, and E. Kamar, “Teaching language models to hallucinate less with synthetic tasks,”arXiv preprint arXiv:2310.06827, 2023

2023 arXiv
[191]

On the universal truthfulness hyperplane inside llms,

J. Liu, S. Chen, Y . Cheng, and J. He, “On the universal truthfulness hyperplane inside llms,”arXiv preprint arXiv:2407.08582, 2024

2024 arXiv
[192]

Halu-NLP at SemEval-2024 task 6: MetaCheckGPT - a multi-task hallucination detection using LLM uncertainty and meta-models,

R. Mehta, A. Hoblitzell, J. O’keefe, H. Jang, and V . Varma, “Halu-NLP at SemEval-2024 task 6: MetaCheckGPT - a multi-task hallucination detection using LLM uncertainty and meta-models,” inProceedings of the 18th International Workshop on Semantic Evaluation (SemEval- 2024), A...

2024
[193]

Truth forest: Toward multi-scale truthfulness in large language models through intervention without tuning,

Z. Chen, X. Sun, X. Jiao, F. Lian, Z. Kang, D. Wang, and C. Xu, “Truth forest: Toward multi-scale truthfulness in large language models through intervention without tuning,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 19, 2024, pp. 20 967– 20 974

2024
[194]

Improving factuality in large language models via decoding-time hal- lucinatory and truthful comparators,

D. Yang, D. Xiao, J. Wei, M. Li, Z. Chen, K. Li, and L. Zhang, “Improving factuality in large language models via decoding-time hal- lucinatory and truthful comparators,”arXiv preprint arXiv:2408.12325, 2024

2024 arXiv
[195]

Decoder tuning: Efficient language understanding as decoding,

G. Cui, W. Li, N. Ding, L. Huang, Z. Liu, and M. Sun, “Decoder tuning: Efficient language understanding as decoding,”arXiv preprint arXiv:2212.08408, 2022

2022 arXiv
[196]

Hallucination detection in large language models with metamorphic relations,

B. Yang, M. A. A. Mamun, J. M. Zhang, and G. Uddin, “Hallucination detection in large language models with metamorphic relations,” 2025. [Online]. Available: https://arxiv.org/abs/2502.15844

2025 arXiv
[197]

Lynx: An open source hallucination evaluation model,

S. S. Ravi, B. Mielczarek, A. Kannappan, D. Kiela, and R. Qian, “Lynx: An open source hallucination evaluation model,” 2024. [Online]. Available: https://arxiv.org/abs/2407.08488

2024 arXiv
[198]

Llms instead of human judges? a large scale empirical 31 study across 20 nlp evaluation tasks,

A. Bavaresco, R. Bernardi, L. Bertolazzi, D. Elliott, R. Fern ´andez, A. Gatt, E. Ghaleb, M. Giulianelli, M. Hanna, A. Koller, A. F. T. Martins, P. Mondorf, V . Neplenbroek, S. Pezzelle, B. Plank, D. Schlangen, A. Suglia, A. K. Surikuchi, E. Takmaz, and A. Testoni, “Llms inste...

2024 arXiv
[199]

Large legal fictions: Profiling legal hallucinations in large language models,

M. Dahl, V . Magesh, M. Suzgun, and D. E. Ho, “Large legal fictions: Profiling legal hallucinations in large language models,”Journal of Legal Analysis, vol. 16, no. 1, pp. 64–93, 2024

2024
[200]

Medhallbench: A new benchmark for assessing hallucination in medical large language models,

K. Zuo and Y . Jiang, “Medhallbench: A new benchmark for assessing hallucination in medical large language models,” 2025. [Online]. Available: https://arxiv.org/abs/2412.18947

2025 arXiv
[201]

An open-source data contamination report for large language models,

Y . Li, Y . Guo, F. Guerin, and C. Lin, “An open-source data contamination report for large language models,” inFindings of the Association for Computational Linguistics: EMNLP 2024, Y . Al- Onaizan, M. Bansal, and Y .-N. Chen, Eds. Miami, Florida, USA: Association for Computa...

2024
[202]

Thinkbench: Dynamic out-of-distribution evaluation for robust llm reasoning,

S. Huang, L. Yang, Y . Song, S. Chen, L. Cui, Z. Wan, Q. Zeng, Y . Wen, K. Shao, W. Zhang, J. Wang, and Y . Zhang, “Thinkbench: Dynamic out-of-distribution evaluation for robust llm reasoning,”
[203]

Unsupervised real-time hallucination detection based on the internal states of large language models,

W. Su, C. Wang, Q. Ai, Y . HU, Z. Wu, Y . Zhou, and Y . Liu, “Unsupervised real-time hallucination detection based on the internal states of large language models,” 2024. [Online]. Available: https://arxiv.org/abs/2403.06448

2024 arXiv
[204]

Benchmark self-evolving: A multi-agent framework for dynamic llm evaluation,

S. Wang, Z. Long, Z. Fan, Z. Wei, and X. Huang, “Benchmark self-evolving: A multi-agent framework for dynamic llm evaluation,”
[205]

Dynamic evaluation of large language models by meta probing agents,

K. Zhu, J. Wang, Q. Zhao, R. Xu, and X. Xie, “Dynamic evaluation of large language models by meta probing agents,” 2024. [Online]. Available: https://arxiv.org/abs/2402.14865

2024 arXiv
[206]

(2023) Improve llm performance with human and ai feedback on amazon sagemaker for amazon engineering

Amazon Web Services. (2023) Improve llm performance with human and ai feedback on amazon sagemaker for amazon engineering. [Online]. Available: https://aws.amazon.com/cn/blogs/machine- learning/improve-llm-performance-with-human-and-ai-feedback-on- amazon-sagemaker-for-amazon-...

2023
[207]

Diaz and M

F. Diaz and M. Madaio,Scaling Laws Do Not Scale. AAAI Press, 2025, p. 341–357

2025
[208]

Available: https://arxiv.org/abs/2402.11443

[Online]. Available: https://arxiv.org/abs/2402.11443

arXiv
[209]

Lima: Less is more for alignment,

C. Zhou, P. Liu, P. Xu, S. Iyer, J. Sun, Y . Mao, X. Ma, A. Efrat, P. Yu, L. Yuet al., “Lima: Less is more for alignment,”Advances in Neural Information Processing Systems, vol. 36, 2024

2024
[210]

Alignment at pre-training! towards native alignment for arabic LLMs,

J. Liang, Z. Cai, J. Zhu, H. Huang, K. Zong, B. An, M. Alharthi, J. He, L. Zhang, H. Li, B. Wang, and J. Xu, “Alignment at pre-training! towards native alignment for arabic LLMs,” inThe Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024. [Online]. A...

2024
[211]

Saullm- 54b & saullm-141b: Scaling up domain adaptation for the legal domain,

P. Colombo, T. Pessoa Pires, M. Boudiaf, R. Melo, G. Hautreux, E. Malaboeuf, J. Charpentier, D. Culver, and M. Desa, “Saullm- 54b & saullm-141b: Scaling up domain adaptation for the legal domain,” inAdvances in Neural Information Processing Systems, A. Globerson, L. Mackey...

2024
[212]

Improving data efficiency via curating llm-driven rating systems,

J. Pang, J. Wei, A. P. Shah, Z. Zhu, Y . Wang, C. Qian, Y . Liu, Y . Bao, and W. Wei, “Improving data efficiency via curating llm-driven rating systems,” 2024. [Online]. Available: https://arxiv.org/abs/2410.10877

2024 arXiv
[213]

Spiral of silences: How is large language model killing information retrieval?–a case study on open domain question answering,

X. Chen, B. He, H. Lin, X. Han, T. Wang, B. Cao, L. Sun, and Y . Sun, “Spiral of silences: How is large language model killing information retrieval?–a case study on open domain question answering,”arXiv preprint arXiv:2404.10496, 2024

2024 arXiv
[216]

Medical adaptation of large language and vision-language models: Are we making progress?

D. P. Jeong, S. Garg, Z. C. Lipton, and M. Oberst, “Medical adaptation of large language and vision-language models: Are we making progress?” 2024. [Online]. Available: https://arxiv.org/abs/2411.04118

2024 arXiv
[2023]

Available: https://arxiv.org/abs/2303.08774

[Online]. Available: https://arxiv.org/abs/2303.08774

arXiv
[2024]

Available: https://arxiv.org/abs/2407.04693

[Online]. Available: https://arxiv.org/abs/2407.04693

arXiv
[2025]

Available: https://arxiv.org/abs/2502.16268

[Online]. Available: https://arxiv.org/abs/2502.16268

arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [23]

Hallucination is inevitable: An innate limitation of large language models,

A. Author and B. Coauthor, “Hallucination is inevitable: An innate limitation of large language models,”Journal of Machine Learning Research, vol. 25, no. 1, pp. 1–30, 2024, theoretical proof that LLM hallucinations arise from fundamental trade-offs in autoregressive generation. [Online]. Available: https://jmlr.org/papers/v25/author2024.html

2024

[2] [1]

Recent advances in natural language processing via large pre-trained language models: A survey,

B. Min, H. Ross, E. Sulem, A. P. B. Veyseh, T. H. Nguyen, O. Sainz, E. Agirre, I. Heintz, and D. Roth, “Recent advances in natural language processing via large pre-trained language models: A survey,”ACM Computing Surveys, vol. 56, no. 2, pp. 1–40, 2023

2023

[3] [3]

Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,

D. Guo, D. Yang, H. Zhang, J. Song, R. Zhang, R. Xu, Q. Zhu, S. Ma, P. Wang, X. Biet al., “Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,”arXiv preprint arXiv:2501.12948, 2025

arXiv 2025

[4] [4]

A comprehensive survey of hallucination mitigation tech- niques in large language models,

S. Tonmoy, S. Zaman, V . Jain, A. Rani, V . Rawte, A. Chadha, and A. Das, “A comprehensive survey of hallucination mitigation tech- niques in large language models,”arXiv preprint arXiv:2401.01313, 2024

arXiv 2024

[5] [5]

hallucination,

“hallucination,” https://www.etymonline.com/word/hallucination, ac- cessed: 2025-03-03

2025

[6] [6]

Des maladies mentales considerees sous les rapports medical, hygi ´enique et medico-legal,

J.-E. D. Esquirol, “Des maladies mentales considerees sous les rapports medical, hygi ´enique et medico-legal,”Bailliere, 1838

[7] [7]

Hallucinations in schizophrenia and parkinson’s disease,

G. F ´enelon, F. Mahieux, R. Huon, and M. Ziegler, “Hallucinations in schizophrenia and parkinson’s disease,”Psychopathology, vol. 33, no. 6, pp. 299–306, 2000. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/11124598/

arXiv 2000

[8] [8]

Six challenges for neural machine transla- tion,

P. Koehn and R. Knowles, “Six challenges for neural machine transla- tion,”arXiv preprint arXiv:1706.03872, 2017

arXiv 2017