Pith. sign in

REVIEW 4 major objections 6 minor 36 references

Are the Hidden States Hiding Something? Testing the Limits of Factuality-Encoding Capabilities in LLMs

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Hidden-state factuality probes trained on templated sentences fail to classify facts generated by LLMs themselves.

desk verdict A mostly solid reproduction plus a genuinely new negative result, undercut by a dataset filter that selects only the model's hardest, most ambiguous facts. read the letter →

arxiv 2505.16520 v3 pith:5XI6PVB2 submitted 2025-05-22 cs.CL cs.AI

Giovanni Servedio , Alessandro De Bellis , Dario Di Palma , Vito Walter Anelli , Tommaso Di Noia This is my paper
classification cs.CLcs.AI
keywords factualityencodinghiddenstatesprobingclassifiershallucinationdetectionself-evaluationLLM-generatedtexttruthfulnessperplexity-basednegativesampling
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 asks whether a language model's internal activations really encode whether a statement is true, as earlier work claimed, and whether that signal survives when statements come from the model's own generative distribution. It first reproduces the SAPLMA probing study on templated true-false sentences and confirms the original results. It then introduces two harder dataset-construction strategies: false statements sampled to be plausible to the model, and true-false facts generated by the model itself from question-answering collections. The central finding is that probes that do well on templated facts fall to near chance on LLM-generated facts from TriviaQA, SQuAD 2.0, and TruthfulQA. The reader should care because it defines the boundary of hidden-state truthfulness: internal factuality signals may exist, but they do not yet transfer to the realistic case of judging the model's own outputs.

What carries the argument

The carrying object is the SAPLMA probe, a three-layer feedforward classifier trained on hidden-state activations at selected transformer layers and originally proposed by Azaria and Mitchell (2023). The paper builds two new dataset mechanisms around it. The first is perplexity-guided negative sampling, which selects false statements by weighting alternative property values through the target LLM's own token likelihoods, so false sentences are plausible to the model. The second is an LLM-generated fact pipeline that prompts the model multiple times with questions from QA collections, keeps only questions whose correct-answer ratio is near one half, and labels the generated answers with an LLM judge. These mechanisms isolate what changes when evaluation moves from templated statements to model-generated text.

What would settle it

A replication in which the same probes are trained and tested on an LLM-generated fact dataset whose labels are verified by human annotators, and any layer reaches clearly above-chance accuracy (for example above 0.65), would contradict the paper's central conclusion.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the factuality-encoding result of Azaria and Mitchell is reproducible on synthetic templated statements but does not carry over to LLM-generated facts. Using a probe trained and evaluated on sentences produced by Llama 2-7b and OPT-6.7b answering TriviaQA questions, accuracy hovers near chance across layers even after threshold tuning, and the same pattern holds for SQuAD 2.0 and TruthfulQA. The authors interpret this as showing that hidden states carry some factuality signal, enough to separate easy true-false templates, but not a general task-independent signal that lets a probe judge whatever the model itself generates.

Load-bearing premise

The load-bearing premise is that the GPT-4o-mini judge's true and false labels for the LLM-generated answers are accurate enough to serve as ground truth; if those labels are biased or noisy, the near-chance probe results cannot be cleanly interpreted.

Editorial extensions

If this is right

  • SAPLMA-style probing cannot yet serve as a reliable factuality self-evaluation tool for model-generated text, because near-chance accuracy on open-domain QA facts would make abstention or self-correction driven by such probes no better than guessing.
  • High accuracy on templated true-false datasets is not evidence that an LLM knows when it is lying in general, so evaluations of factuality encoding must include statements drawn from the model's own generative distribution.
  • Perplexity-based negative sampling produces harder, more plausible false statements and lowers the average perplexity gap between true and false sentences, yet probe accuracy stays comparable, indicating the difficulty comes from the data distribution rather than statement plausibility alone.
  • Future factuality probes should be trained and tested on LLM-generated facts, and may need uncertainty-aware classification or other signals beyond a fixed hidden-state probe.

Reading between the lines

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

  • Editorial inference: if the result holds in larger models, it casts doubt on hidden-state probing as a standalone hallucination detector and points toward combining probes with output-level self-consistency or retrieval verification.
  • Editorial inference: the near-chance probe accuracy could partly reflect the LLM judge's labels rather than an absent factuality signal; a human-annotated subset would separate 'the signal is not in hidden states' from 'the labels do not match what the model was saying'.
  • Editorial inference: the perplexity-guided negative-sampling strategy is reusable as a benchmark generator that produces hard negatives plausible to the model, which could make other truthfulness evaluations more stringent.
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 / 6 minor

Summary. The paper re-examines whether hidden states of LLMs linearly encode statement factuality, in the setting of the SAPLMA probe of Azaria and Mitchell (2023). It first reproduces the original experiments on Llama 2-7b and OPT-6.7b, reporting close agreement with the published accuracy values (Tables 1-3). It then proposes two dataset-construction strategies: a perplexity-based negative-sampling procedure that generates plausible false statements from tabular data (Section 4.1), and an LLM-generated fact set built by sampling answers to TriviaQA, SQuAD 2.0, and TruthfulQA questions, with GPT-4o mini as a veracity oracle (Section 4.2). The experiments address RQ1 (probes trained on the original versus the refined synthetic data transfer comparably to the refined test set) and RQ2 (probes trained and tested on LLM-generated facts stay near chance, with best average accuracy about 0.59 in Table 6). The paper concludes that prior findings are only partially validated and that generalization to LLM-generated datasets remains challenging. Code is released.

Significance. If established as stated, the main negative result would be a useful contribution to the factuality-probing literature, complementing recent findings by Orgad et al. (2024) and Levinstein and Herrmann (2024) that factual-signal decodability is dataset- and task-dependent. The paper's concrete strengths are the careful reproduction protocol (five layers, two models, 100 probes), the public code release, the matched-perplexity design of the refined synthetic dataset, and the consistency of the RQ2 outcome across three QA collections and two open models. The principal limitation is scope: the negative result is demonstrated only on the p approximately 0.5 subset selected by Eq. (4), and the missing control experiments mean the claim must currently be read as conditional rather than as a general failure of factuality probing on LLM-generated content.

major comments (4)
  1. [Section 4.2, Eq. (4), and the claims of Section 5.3 and the Abstract.] The central negative claim of the paper is established only for the subset of questions selected by the filter in Eq. (4), |p_i - 0.5| < tau with tau = 0.1, i.e., questions for which the target model produces roughly as many correct as incorrect answers over K = 10 generations. This deliberately isolates the model's decision boundary, the regime in which a factuality signal that scales with knowledge strength or confidence is hardest to decode, and in the train-on-TriviaQA condition of Table 6 the probe is trained exclusively on such boundary items, so it never observes the input regime in which the original SAPLMA results were obtained. Near-chance accuracy on this subset is therefore at least partly confounded with the difficulty of the selected items and cannot, by itself, contradict the hypothesis that factuality is linearly decodable for confident model generations. Yet the abstract, Section 5.3, and the Conclusion state the failure as a claim about 'LLM-generated datasets' in general, without this qualification. Because the same filter is applied to SQuAD 2.0 and TruthfulQA in Appendix E, the cross-dataset consistency does not resolve the scope problem. A control experiment without the filter or with progressively larger tau (e.g., tau = 0.3, 0.5, no filtering, or a matched subset with p far from 0.5) is needed to determine whether the negative result extends to typical LLM-generated content; without it the headline claim is not yet established.
  2. [Section 4.2, Eq. (3), Appendix C.2, and the Limitations section.] The veracity labels that drive both the Eq. (4) filter and the probe targets are produced by a single oracle model, GPT-4o mini at temperature 1 with a 3-shot prompt, and no human validation is reported; the paper's Limitations section correctly asks for manual verification, but the analysis proceeds as if the labels were noise-free. Since the same labels determine which questions enter the dataset and what the probe is trained and evaluated on, systematic oracle bias (e.g., consistent mislabeling of a particular answer style) would distort both the composition of the p approximately 0.5 subset and the measured probe accuracy, making the near-chance result hard to interpret as evidence about hidden-state factuality encoding. Some labels in Table 9 are non-obvious (for instance, 'Anakin Skywalker is Darth Vader's son' being labeled false), which underscores the need for a verification step. Reporting the oracle's agreement with human annotations on a sample and checking the sensitivity of Table 6 to the oracle choice would substantially strengthen the RQ2 conclusion.
  3. [The point estimates in Tables 5, 6, 13, and 16.] All accuracy figures in the paper are point estimates without variance, confidence intervals, or significance tests. This omission matters most for the central comparison, because the values in Table 6 hover near chance and individual cells fall below 0.5 (e.g., triviacountry OPT optimal-threshold 0.449, triviabug Llama last-layer 0.494), which is compatible with substantial run-to-run variation even if the averages are nominally above chance given the large aggregated sample sizes. The reproduction section already trains 20 probes per layer (Section 2.3), so reporting the dispersion of the probe accuracies (or at least bootstrap intervals) for the novel settings is a natural addition; without it, the statement in Section 5.3 that accuracy 'does not reach a noteworthy threshold' is not backed by a quantified comparison.
  4. [Section 5.2 and the design of Table 5, in relation to RQ1.] The RQ1 conclusion that the probes 'can generalize even when the train-test datasets have different perplexity' is not fully supported by the experimental design. Table 5 varies the training dataset (original versus refined) while keeping the test set fixed to the refined dataset, so it measures robustness to a change in training distribution, not generalization across test distributions; the complementary cell, training on the refined dataset and testing on the original dataset, is missing. The accompanying claim that results are 'mostly comparable' is also uneven: for Llama 2-7b the original-trained probes outperform the refined-trained probes by about 0.06 at layer 28 (0.666 versus 0.609) and 0.07 at layer 24 (0.708 versus 0.642), and the Cities topic drops from 0.706 to 0.509 at layer 28 and from 0.829 to 0.603 at layer 24, so the wording overstates the level of agreement and the dropped cells deserve discussion.
minor comments (6)
  1. [Section 5.1 and Appendix B.] Section 5.1 points to Appendix B for details on excluding the Scientific Facts topic, but Appendix B discusses only the Cities-template restriction and does not describe that exclusion; the pointer should be corrected or the discussion added, since the exclusion changes the topic set in Table 5 from six to five topics and affects comparability with Tables 1 and 2.
  2. [Section 4.1 and the Limitations section.] The perplexity-based pipeline of Section 4.1 assumes that a low-perplexity true statement indicates model knowledge (the alpha-filter) and that perplexity is a valid plausibility proxy, an assumption the paper itself calls into question in the Limitations by citing Duan et al. (2024); one sentence in Section 5.2 acknowledging how this affects the interpretation of Table 5 would help calibrate the RQ1 answer.
  3. [Table 6 and Table 8.] Table 6 reports per-source accuracies over very different sample sizes (97 to 1054 sentences per source in Table 8), and the reader must cross-reference the two tables to know which averages are reliable; adding the sentence counts to Table 6, or a note in the caption, would improve readability.
  4. [Eq. (2) and Section 5.1.] The probability distribution in Eq. (2) is typeset ambiguously; writing the denominator explicitly as the sum over candidates in C* would remove the ambiguity, and the relationship between the sampling parameters k = 10 and p = 0.9 (Section 5.1) and the threshold parameters alpha and beta should be stated once in a single place.
  5. [Tables 6 and 8.] The source name 'wwwodquiz' in Table 8 appears as 'odquiz' in Table 6; the two should be aligned.
  6. [Section 3.1 and Table 2.] In Section 3.1 the comparison of the Table 2 baselines with SAPLMA requires cross-referencing Table 1, since the key point is that all baselines are near chance (0.537-0.559) while SAPLMA reaches about 0.65-0.83 on the same data; restating those two reference points in the text would make the reproduction claims easier to verify.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central negative result is benchmarked against externally sourced QA data, and the p≈0.5 filter is a dataset-selection choice rather than an input-equivalent derivation.

full rationale

The paper's derivation chain does not reduce to its own inputs. The reproduction of Azaria and Mitchell (2023) is checked against the original released dataset and published numbers, and the observed agreement is an independent empirical outcome rather than a consequence of the paper's definitions. The perplexity-based dataset construction uses the target LLM's perplexity to select and sample plausible false statements, but the training labels remain the external true/false annotations from the original tabular sources; the probe's accuracy on this refined dataset is therefore not forced by the construction. The LLM-generated dataset uses externally sourced questions from TriviaQA, SQuAD 2.0, and TruthfulQA, with veracity labels assigned by GPT-4o mini as an oracle; the near-chance probe accuracy is measured against those oracle labels and is not a renamed fit parameter. The Eq. 4 filter |p_i − 0.5| < τ does restrict evaluation to questions where the target model is internally inconsistent across K = 10 generations, which is a legitimate threat to the external validity of the negative claim: the conclusion that probes fail on 'LLM-generated datasets' is only directly supported for this deliberately ambiguous subset. However, this is a sampling-design limitation, not circularity, because the probe's chance-level performance does not follow by construction from the filter; it remains an empirical finding about that subset. Self-citations appear only as background context in the introduction and related work and are not load-bearing for any central claim. No equation, fitted parameter, or self-citation chain makes the predictions equivalent to the inputs. The appropriate finding is therefore no significant circularity.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The paper's central claims rest on a small set of hand-picked hyperparameters (alpha, beta, k, p, tau, K), two domain assumptions about what perplexity and an oracle judge measure, and two ad hoc dataset restrictions.

free parameters (7)
  • alpha (true-statement knowledge filter) = 0.1
    In Section 4.1, alpha sets the fraction k = alpha*|C| of lowest-perplexity candidates required for a true statement to be retained; no sensitivity analysis is provided.
  • beta (false-candidate perplexity margin) = 0.1
    Section 4.1: false candidates are retained only if PP(c) < (1+beta) PP(true); chosen by hand.
  • top-k (k) = 10
    Section 5.1: top-k sampling parameter for selecting false statement candidates.
  • nucleus probability (p) = 0.9
    Section 5.1: nucleus sampling threshold applied after top-k selection.
  • tolerance tau = 0.1
    Section 5.1: selects only questions where the model's correct-answer rate is within 0.1 of 0.5, shaping the LLM-generated dataset.
  • number of samples K = 10
    Section 5.1: each question is answered K times to estimate the correct-answer ratio p_i.
  • minimum answer length = 5 tokens
    Section 5.1: answers shorter than 5 tokens are discarded for self-consistency.
assumptions (6)
  • domain assumption Perplexity of a sentence is a valid proxy for how plausible that sentence is to the LLM and whether the LLM 'knows' the fact.
    Section 4.1 uses perplexity to filter true statements the model supposedly knows and to weight false candidates by plausibility. The Limitations section itself notes Duan et al. (2024) found LLMs assign similar perplexity to seen and unseen sentences, which undermines this assumption.
  • domain assumption The oracle LLM (GPT-4o mini) provides correct veracity labels for LLM-generated answers given the ground-truth answer and question.
    Section 4.2 Eq. (3) and Appendix C.2 treat the judge model's 0/1 labels as ground truth; the paper only suggests human verification as future work, so the main RQ2 result rests on this.
  • domain assumption An LLM can encode factuality about a generated fact only if it has some prior knowledge of that fact, and this knowledge is captured by a mixed correct/incorrect response pattern (p_i near 0.5).
    Section 4.2, paragraph before Eq. (4), justifies restricting the dataset to questions where the model is correct about half the time.
  • domain assumption Hidden states extracted from the last token are the appropriate representation for factuality probing.
    Section 2.3: the authors identified the last token from code inspection of Azaria and Mitchell's work; the probing architecture inherits this choice without validation.
  • ad hoc to paper The Scientific Facts topic can be excluded from the perplexity-based sampling because it was originally generated by ChatGPT, not from tabular data.
    Section 5.1 and Appendix B exclude this topic, reducing the original six-topic setup to five; this changes the evaluation domain without a principled argument that the exclusion does not bias conclusions.
  • ad hoc to paper For the Cities topic, restricting to the template '<city> is a city in <country>' and discarding binary-property templates does not distort the factuality assessment.
    Appendix B narrows the Cities facts to a single template to avoid easy-to-classify samples, a post-hoc modeling choice that affects dataset difficulty.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Are the Hidden States Hiding Something? Testing the Limits of Factuality-Encoding Capabilities in LLMs." pith.science (2026). https://pith.science/paper/5XI6PVB2

@misc{pith2026250516520,
  author       = {Pith},
  title        = {Pith review of: Are the Hidden States Hiding Something? Testing the Limits of Factuality-Encoding Capabilities in LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5XI6PVB2}},
  note         = {Machine review of arXiv:2505.16520}
}
read the original abstract

Factual hallucinations are a major challenge for Large Language Models (LLMs). They undermine reliability and user trust by generating inaccurate or fabricated content. Recent studies suggest that when generating false statements, the internal states of LLMs encode information about truthfulness. However, these studies often rely on synthetic datasets that lack realism, which limits generalization when evaluating the factual accuracy of text generated by the model itself. In this paper, we challenge the findings of previous work by investigating truthfulness encoding capabilities, leading to the generation of a more realistic and challenging dataset. Specifically, we extend previous work by introducing: (1) a strategy for sampling plausible true-false factoid sentences from tabular data and (2) a procedure for generating realistic, LLM-dependent true-false datasets from Question Answering collections. Our analysis of two open-source LLMs reveals that while the findings from previous studies are partially validated, generalization to LLM-generated datasets remains challenging. This study lays the groundwork for future research on factuality in LLMs and offers practical guidelines for more effective evaluation.

Figures

Figures reproduced from arXiv: 2505.16520 by the authors.

Figure 1
Figure 1. Comparison of average perplexity scores for [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 10 canonical work pages

  1. [1]

    Guillaume Alain. 2016. Understanding intermediate layers using linear classifier probes. arXiv preprint arXiv:1610.01644

    arXiv 2016

  2. [2]

    Vito Walter Anelli, Giovanni Maria Biancofiore, Alessandro De Bellis, Tommaso Di Noia, and Eugenio Di Sciascio. 2022. https://doi.org/10.1145/3511808.3557617 Interpretability of BERT latent space through knowledge graphs . In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, Atlanta, GA, USA, October 17-21, 2022 ,...

    arXiv 2022

  3. [3]

    Mitchell

    Amos Azaria and Tom M. Mitchell. 2023. The internal state of an LLM knows when it's lying. In EMNLP (Findings) , pages 967--976. Association for Computational Linguistics

    work page 2023

  4. [4]

    Giovanni Maria Biancofiore, Dario Di Palma, Claudio Pomo, Fedelucio Narducci, and Tommaso Di Noia. 2025. Conversational user interfaces and agents. In Human-Centered AI: An Illustrated Scientific Quest, pages 399--438. Springer

    work page 2025

  5. [5]

    Hamprecht, and Boaz Nadler

    Lennart B \" u rger, Fred A. Hamprecht, and Boaz Nadler. 2024. Truth is universal: Robust detection of lies in llms. In NeurIPS

    work page 2024

  6. [6]

    Nitay Calderon, Roi Reichart, and Rotem Dror. 2025. The alternative annotator test for llm-as-a-judge: How to statistically justify replacing human annotators with llms. CoRR, abs/2501.10970

    arXiv 2025

  7. [7]

    Chao Chen, Kai Liu, Ze Chen, Yi Gu, Yue Wu, Mingyuan Tao, Zhihang Fu, and Jieping Ye. 2024. Inside: Llms' internal states retain the power of hallucination detection. arXiv preprint arXiv:2402.03744

    arXiv 2024

  8. [8]

    Alexis Conneau, German Kruszewski, Guillaume Lample, Lo \" c Barrault, and Marco Baroni. 2018. What you can cram into a single vector: Probing sentence embeddings for linguistic properties. arXiv preprint arXiv:1805.01070

    arXiv 2018

Show all 36 references
  1. [9]

    Matthew Dahl, Varun Magesh, Mirac Suzgun, and Daniel E. Ho. 2024. Large legal fictions: Profiling legal hallucinations in large language models. CoRR, abs/2401.01301

    2024 arXiv

  2. [10]

    Alessandro De Bellis , Vito Walter Anelli, Tommaso Di Noia, and Eugenio Di Sciascio. 2024. PRONTO: prompt-based detection of semantic containment patterns in MLM s. In ISWC (2) , volume 15232 of Lecture Notes in Computer Science, pages 227--246. Springer

    2024

  3. [11]

    Dario Di Palma . 2023. Retrieval-augmented recommender system: Enhancing recommender systems with large language models. In RecSys, pages 1369--1373. ACM

    2023

  4. [12]

    Dario Di Palma, Felice Antonio Merra, Maurizio Sfilio, Vito Walter Anelli, Fedelucio Narducci, and Tommaso Di Noia. 2025. https://doi.org/10.1145/3726302.3730178 Do llms memorize recommendation datasets? a preliminary study on movielens-1m . In Proceedings of the 48th Internat...

    2025

  5. [13]

    Michael Duan, Anshuman Suri, Niloofar Mireshghallah, Sewon Min, Weijia Shi, Luke Zettlemoyer, Yulia Tsvetkov, Yejin Choi, David Evans, and Hannaneh Hajishirzi. 2024. Do membership inference attacks work on large language models? CoRR, abs/2402.07841

    2024 arXiv

  6. [14]

    Shangbin Feng, Weijia Shi, Yike Wang, Wenxuan Ding, Vidhisha Balachandran, and Yulia Tsvetkov. 2024. https://doi.org/10.18653/v1/2024.acl-long.786 Don`t hallucinate, abstain: Identifying LLM knowledge gaps via multi- LLM collaboration . In Proceedings of the 62nd Annual Meetin...

    2024 doi

  7. [15]

    Zorik Gekhman, Eyal Ben - David, Hadas Orgad, Eran Ofek, Yonatan Belinkov, Idan Szpektor, Jonathan Herzig, and Roi Reichart. 2025. Inside-out: Hidden factual knowledge in llms. CoRR, abs/2503.15299

    2025 arXiv

  8. [16]

    Jiawei Gu, Xuhui Jiang, Zhichao Shi, Hexiang Tan, Xuehao Zhai, Chengjin Xu, Wei Li, Yinghan Shen, Shengjie Ma, Honghao Liu, Yuanzhuo Wang, and Jian Guo. 2024. A survey on llm-as-a-judge. CoRR, abs/2411.15594

    2024 arXiv

  9. [17]

    Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. 2020. https://openreview.net/forum?id=rygGQyrFvH The curious case of neural text degeneration . In 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020 . Ope...

    2020

  10. [18]

    Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, and Ting Liu. 2024. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Trans. Inf. Syst

    2024

  11. [19]

    Ziwei Ji, Tiezheng Yu, Yan Xu, Nayeon Lee, Etsuko Ishii, and Pascale Fung. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.123 Towards mitigating LLM hallucination via self reflection . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 1827-...

    2023 doi

  12. [20]

    Weld, and Luke Zettlemoyer

    Mandar Joshi, Eunsol Choi, Daniel S. Weld, and Luke Zettlemoyer. 2017. Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension. In ACL (1) , pages 1601--1611. Association for Computational Linguistics

    2017

  13. [21]

    Saurav Kadavath, Tom Conerly, Amanda Askell, Tom Henighan, Dawn Drain, Ethan Perez, Nicholas Schiefer, Zac Hatfield - Dodds, Nova DasSarma, Eli Tran - Johnson, Scott Johnston, Sheer El Showk, Andy Jones, Nelson Elhage, Tristan Hume, Anna Chen, Yuntao Bai, Sam Bowman, Stanislav...

    2022 arXiv

  14. [22]

    Benjamin A Levinstein and Daniel A Herrmann. 2024. Still no lie detector for language models: Probing empirical and conceptual roadblocks. Philosophical Studies, pages 1--27

    2024

  15. [23]

    Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. Truthfulqa: Measuring how models mimic human falsehoods. In ACL (1) , pages 3214--3252. Association for Computational Linguistics

    2022

  16. [24]

    Samuel Marks and Max Tegmark. 2023. The geometry of truth: Emergent linear structure in large language model representations of true/false datasets. CoRR, abs/2310.06824

    2023 arXiv

  17. [25]

    Hadas Orgad, Michael Toker, Zorik Gekhman, Roi Reichart, Idan Szpektor, Hadas Kotek, and Yonatan Belinkov. 2024. Llms know more than they show: On the intrinsic representation of LLM hallucinations. CoRR, abs/2410.02707

    2024 arXiv

  18. [26]

    Duy Khoa Pham and Bao Quoc Vo. 2024. Towards reliable medical question answering: Techniques and challenges in mitigating hallucinations in language models. CoRR, abs/2408.13808

    2024 arXiv

  19. [27]

    Pranav Rajpurkar, Robin Jia, and Percy Liang. 2018. Know what you don't know: Unanswerable questions for squad. In ACL (2) , pages 784--789. Association for Computational Linguistics

    2018

  20. [28]

    Ian Tenney, Patrick Xia, Berlin Chen, Alex Wang, Adam Poliak, R Thomas McCoy, Najoung Kim, Benjamin Van Durme, Samuel R Bowman, Dipanjan Das, et al. 2019. What do you learn from context? probing for sentence structure in contextualized word representations. arXiv preprint arXi...

    2019 arXiv

  21. [29]

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, and Yasmine Babaei et al. 2023. Llama 2: Open foundation and fine-tuned chat models. CoRR, abs/2307.09288

    2023 arXiv

  22. [30]

    Shriyash Upadhyay, Etan Ginsberg, and Chris Callison - Burch. 2023. Improving mathematics tutoring with A code scratchpad. In BEA@ACL, pages 20--28. Association for Computational Linguistics

    2023

  23. [31]

    David Wadden, Shanchuan Lin, Kyle Lo, Lucy Lu Wang, Madeleine van Zuylen, Arman Cohan, and Hannaneh Hajishirzi. 2020. https://doi.org/10.18653/v1/2020.emnlp-main.609 Fact or fiction: Verifying scientific claims . In Proceedings of the 2020 Conference on Empirical Methods in Na...

    2020 doi

  24. [32]

    Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, et al. 2022. Opt: Open pre-trained transformer language models. arXiv preprint arXiv:2205.01068

    2022 arXiv

  25. [33]

    Xiaokang Zhang, Zijun Yao, Jing Zhang, Kaifeng Yun, Jifan Yu, Juanzi Li, and Jie Tang. 2024. Transferable and efficient non-factual content detection via probe training with offline consistency checking. arXiv preprint arXiv:2404.06742

    2024 arXiv

  26. [34]

    Yue Zhang, Yafu Li, Leyang Cui, Deng Cai, Lemao Liu, Tingchen Fu, Xinting Huang, Enbo Zhao, Yu Zhang, Yulong Chen, et al. 2023. Siren's song in the ai ocean: a survey on hallucination in large language models. arXiv preprint arXiv:2309.01219

    2023 arXiv

  27. [35]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  28. [36]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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