Pith. sign in

REVIEW 6 major objections 6 minor 30 references

Patchfinder: Leveraging Visual Language Models for Accurate Information Retrieval using Model Uncertainty

T0 review · 6 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read PatchFinder claims that a small vision-language model can reach 94% extraction accuracy on noisy scanned well documents by using its own token-confidence scores to choose the best image patch, outperforming ChatGPT-4o by 18.5 percentage…

desk verdict Useful engineering paper with a plausible method, but evidence for the headline 94% has several fixable gaps: undefined metric, tiny samples, missing baseline, and confidence selection validated only on centered crops. read the letter →

arxiv 2412.02886 v3 pith:LDAI7WG6 submitted 2024-12-03 cs.CV

classification cs.CV
keywords PatchFindervision-languagemodelsinformationextractionscanneddocumentsmodeluncertaintyMaximumSoftmaxProbabilitywellrecordsdocumentunderstanding
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 paper tries to show that a cheap, fine-tuning-free method can extract structured fields from noisy scanned documents as well as or better than much larger commercial models. The proposed algorithm, PatchFinder, crops a scanned page into overlapping patches, asks a small vision-language model to read each patch, and uses a confidence score—the average log-likelihood of the model's most-likely tokens—to pick both the patch size and the single most reliable patch. On 190 historical well records, PatchFinder with the 4.2-billion-parameter Phi-3v reaches 94% accuracy on latitude, longitude, and true vertical depth, outperforming ChatGPT-4o by 18.5 percentage points, and it also beats both baselines on noisy financial statements. The practical payoff would be that organizations with limited compute can automate large-scale digitization of old records without OCR pipelines or fine-tuning.

What carries the argument

The central object is the Patch Confidence score, defined as $PC(m) = \frac{1}{l}\sum_{i=1}^{l} \log p_\theta(x_i^{k^*}|x_{1\cdots i-1}, z^{(m)})$, where $k^*$ is the token with maximum softmax probability and $z^{(m)}$ is the visual-plus-text context of patch $m$. In plainer terms, it is the mean log-likelihood of the most-likely decoded tokens, i.e., an average of Maximum Softmax Probabilities over the generated answer. PatchFinder uses this score in two stages: Patch Size Optimization searches the development data for the patch size that yields high, low-variance confidence (around 22–25% of the image), and Confidence-Based Prediction computes $PC(m)$ for every overlapping patch and returns the prediction of $m^* = \arg\max_m PC(m)$. The mechanism works by cropping the image into a size the model can attend to well, reducing noise while preserving the target field.

What would settle it

A held-out experiment in which correct and incorrect patches have overlapping Patch Confidence distributions would falsify the central claim: for example, if on the 170 test well records the argmax-confidence patch is no more accurate than a randomly chosen patch, the accuracy–confidence correlation breaks down. Concretely, one could take a document where the model hallucinates a plausible number with high average token probability and show that PatchFinder selects that wrong number over the correct one.

Watch

Extended reading notes

Core claim

The paper's central claim is that a vision-language model's own confidence, measured as the average log-likelihood of its maximum-likelihood output tokens, can be used to locate the most reliable region of a noisy scanned document. PatchFinder first uses this Patch Confidence score to select a patch size that keeps the target field inside the model's positional-encoding comfort zone, then evaluates every overlapping patch at that size and returns the prediction of the most confident one. On the authors' 190-document well-record dataset, this lifts Phi-3v's accuracy from 61.8% to 94.0%, surpassing ChatGPT-4o's 75.4%; the same method reaches 53.3% on 30 synthetically noisy financial statements, where ChatGPT-4o scores 36.7% and vanilla Phi-3v scores 0%. The paper presents this as evidence that patching and maximum-softmax-probability-based confidence can improve VLM extraction without fine-tuning.

Load-bearing premise

The whole method rests on the assumption that a higher Patch Confidence score means a more accurate answer; if the model is confidently wrong on a patch, PatchFinder will happily return that wrong answer.

Editorial extensions

If this is right

  • On the 190 well records, optimized PatchFinder raises accuracy from 61.8% (vanilla Phi-3v) and 75.4% (ChatGPT-4o) to 94.0%.
  • The largest gains appear on the hardest document categories: Pennsylvania goes from 66.7% to 93.3%, and New Mexico from 57.6% (ChatGPT-4o) to 93.9%.
  • On 30 synthetically noisy financial statements, PatchFinder reaches 53.3% versus 36.7% for ChatGPT-4o and 0% for vanilla Phi-3v, indicating the method helps most in high-noise settings.
  • On the CORD receipt dataset, PatchFinder's zero-shot accuracy is 77.2–79.0% versus 46.8% for Donut, showing transfer beyond well records.
  • The method requires no fine-tuning and runs on a single laptop with 64 GB of RAM, making it accessible to organizations without large compute budgets.

Reading between the lines

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

  • If Patch Confidence reliably tracks answer correctness, the same score could serve as a rejection rule: documents whose best patch confidence falls below a threshold could be routed to human review, which would be directly useful for the orphan-well cataloging use case.
  • The cropping mechanism implies a general chunking strategy for small-context VLMs: treating image crops like text chunks could let models with short positional-encoding windows handle high-resolution or large-format documents, a consequence the paper states but does not develop.
  • A natural extension the paper leaves implicit is multi-scale aggregation: instead of choosing one patch size, combine predictions across several sizes weighted by Patch Confidence, which could smooth the sensitivity to the 20–25% optimum.
  • The confidence–accuracy correlation is established on 20 development documents without error bars; a direct test would be to report per-category calibration of Patch Confidence against correctness, especially for Old Colorado and New Mexico.
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

6 major / 6 minor

Summary. The paper proposes PatchFinder, a method to improve vision-language-model (VLM) information extraction from noisy scanned documents. PatchFinder defines a Patch Confidence (PC) score as the mean log-likelihood of the maximum-softmax-probability tokens in the VLM's generated sequence, uses this score to choose a patch size on a development set, partitions documents into overlapping patches, and returns the prediction from the highest-confidence patch. Experiments on 190 historical well records report 94% accuracy for optimized PatchFinder with Phi-3v, outperforming ChatGPT-4o (75.4%) and vanilla Phi-3v (61.8%); additional experiments on 30 noisy financial statements and on CORD/FUNSD are reported. The central methodological claim is that PC reliably selects the patch containing the correct answer, and that patch-size optimization further improves accuracy.

Significance. If the central claim holds, the paper offers a simple, fine-tuning-free technique that could make small open-source VLMs competitive with much larger commercial models on noisy scanned documents, with direct relevance to the CATALOG orphan-well initiative and to document-processing practice more broadly. The algorithm is clearly specified (Algorithm 1), and the experiments span multiple datasets, including a known benchmark (CORD) and a synthetic noise condition. The paper also honestly acknowledges limitations around noise dependence and computational overhead. However, the headline accuracy and the confidence-selection mechanism rest on several underspecified and under-validated points, detailed in the major comments.

major comments (6)
  1. [§4.1 and Table 2] The dataset split is internally inconsistent. The text states that 20 documents (10 Colorado, 10 Pennsylvania) are used for development and 170 for testing, but Table 2 reports results for Pennsylvania (93.3%) despite Pennsylvania having only 10 documents total, and the overall 94% figure appears to include the development documents. Please clarify whether all reported accuracies are on the 170-document test set only, or on the full 190 documents, and if Pennsylvania contributes 0 test documents, explain where the 93.3% comes from. This affects every headline number in the abstract and Section 4.3.
  2. [§4.2, Fig. 4, and Algorithm 1] The correlation between Patch Confidence and extraction accuracy is demonstrated only for centered crops, where the target information is kept in the middle of the patch, while Algorithm 1 applies a fixed grid of overlapping patches with no centering. Under grid sampling, the target can be split across patch boundaries, near the edge, or absent from most patches, so the centering experiment does not validate the confidence-selection mechanism actually used. Please add a control that evaluates PatchFinder under grid sampling with a non-confidence-based selection rule (e.g., center patch, random patch, or a fixed patch) and compare it against max-PC selection; alternatively, provide per-patch accuracy and PC statistics on the grid to show that PC identifies the correct patch.
  3. [§3.3, 'Refinement with Output Filtering'] The output filtering step discards patches whose response is non-numerical, but the paper never reports how often this filter triggers, how many documents have all patches filtered, or what the accuracy is without filtering. If the filter removes a large fraction of incorrect candidates, the reported 94% could be an artifact of filtering rather than of Patch Confidence. Please report the discard rate per dataset and the accuracy with and without the filtering rule.
  4. [§4.1, 'evaluation' paragraph] The accuracy metric is not defined precisely. The paper says 'we use accuracy over F1' but does not specify the matching criterion for latitude, longitude, and TVD: exact string match after normalization? tolerance in decimal degrees or feet? How are the different coordinate formats (DMS vs decimal) handled? Without this definition, the reported accuracies are not reproducible and the comparison to ChatGPT-4o is not auditable. Please state the exact correctness rule and, ideally, provide the per-field accuracy.
  5. [Table 2 and §2.3] The strong OCR+Phi-3-mini baseline from the preliminary experiments (Table 1) is not evaluated on the full 190-document dataset. Since the paper motivates the work by arguing that the two-step OCR+LLM approach is costly and error-prone, and since OCR+Phi-3-mini was the best baseline in the preliminary results, omitting this baseline from the main comparison weakens the central claim of outperforming existing approaches. Please add the OCR+Phi-3-mini result (or a similar OCR+LLM baseline) on the same test split.
  6. [Table 2 and §4.3] The per-category samples are very small (Pennsylvania 10 documents, New Mexico 13) and no error bars, confidence intervals, or significance tests are reported. For example, a single misclassification in Pennsylvania changes the reported accuracy by 10 percentage points, so the 93.3% vs 66.7% difference is not shown to be meaningful. Please report exact counts (X/Y per category) and provide confidence intervals or a statistical test.
minor comments (6)
  1. [Abstract and §3.3] The abstract and Section 2 use 'Phi-3v', while Section 3.3 says 'Phi-3.5-Vision model' and Figure 1 says 'Phi-3-mini-128K-instruct'. Please standardize the model name, since these may refer to different model versions.
  2. [Table 1 caption] The caption says '10 documents from Colorado and Pennsylvania datasets' but the table reports separate percentages for each state; please clarify how many documents per state were used in this preliminary benchmark.
  3. [Eq. (2)] The definition of PC as (1/l) Σ log p(...) is the mean per-token log-likelihood, not the log-likelihood of the full sequence. The surrounding text says 'average log-likelihood of the maximum-likelihood sequence,' which is ambiguous; please rephrase to make clear that the average is over token positions.
  4. [§3.2 and Fig. 2] Patch Size Optimization is described as using ground-truth sequences to determine adequate coverage, but Figure 2 and the surrounding text describe a confidence-based curve that does not reference ground truth. Please clarify the role of ground-truth labels in this step and whether the PC curve is computed on documents with known answers.
  5. [Fig. 4] Figure 4 would benefit from error bars or shading and a statement of how many documents and patches underlie each point; the current figure is consistent with the centered-crop protocol but the reader cannot assess variance.
  6. [Throughout] There are several typos and inconsistent capitalizations, e.g., 'dicussing' at the start of Section 3, and 'PatchFinder' vs 'Patchfinder' (especially in Table 3 and Section 5). A careful proofread is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; PatchFinder's confidence-based patch selection is evaluated against held-out ground-truth labels.

full rationale

The paper's derivation chain is self-contained. Patch Confidence (PC) is computed directly from the VLM's token-level Maximum Softmax Probabilities (Eq. 2) and is validated against extraction accuracy on a 20-document development set (Fig. 4); no fitted parameter is renamed as a prediction. The patch size is chosen either by ground-truth coverage in the dev set or by the PC-versus-size trend, which is standard hyperparameter selection. The test accuracy (94%) is measured on held-out documents using exact-field correctness, so the central claim does not reduce to the confidence definition. The self-citations ([6], [28]) only establish the task (extracting TVD, latitude, longitude from well records) and the dataset context; they do not supply any load-bearing mathematical premise, uniqueness theorem, or ansatz. The mismatch between the centered-crop validation of PC and the grid-patch sampling used in Algorithm 1 is a legitimate generalization/robustness concern, but it is not circularity: the PC selection step is still judged against ground-truth accuracy rather than being built into the accuracy definition. Therefore the paper warrants a circularity score of 0.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central result depends on a small number of hand-tuned choices and domain assumptions rather than on a formal derivation. The patch size is fitted to a 20-document development set; the prompt template is handcrafted; the output filtering rule is a post-hoc exclusion; and the accuracy metric is undefined. No new entities are introduced.

free parameters (4)
  • patch_size = 23% of document image (optimized on 20-document development set)
    Selected in Sect. 3.2 based on Patch Confidence curves (Fig. 2) on the development split; used for all test documents.
  • patch_overlap = not specified
    The paper states patches are overlapping but never reports the overlap amount, which affects the number of patches and confidence estimates.
  • output_filtering_rule = reject non-numerical outputs
    Sect. 3.3 describes removing patches with non-numerical responses; the threshold or rule for flagging is not precisely defined, and the discard rate is not reported.
  • prompt_template = handcrafted per dataset with format hints (e.g., latitude examples for Pennsylvania)
    Sect. 3.3 describes prompt engineering with dataset-specific examples; this is a free parameter tuned by the authors and not systematically evaluated.
assumptions (3)
  • standard math Maximum Softmax Probability is a reliable proxy for model confidence
    Relied on in Sect. 3.1 to define Patch Confidence, based on prior work (Hendrycks and Gimpel). The paper acknowledges this is a proxy and validates it only indirectly.
  • domain assumption Target fields (latitude, longitude, TVD) can be captured by a single patch of about 23% of the document
    Assumed in Sect. 3.2 and 4.3 based on the authors' informal analysis of well record layouts. The paper does not analyze how often fields straddle patch boundaries.
  • domain assumption Accuracy is measured as correct/incorrect per field, but the matching criterion is undefined
    Sect. 4.1 states the authors use accuracy over F1 for field extraction, but no tolerance, normalization, or exact-match rule is given. This makes the reported numbers unverifiable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Patchfinder: Leveraging Visual Language Models for Accurate Information Retrieval using Model Uncertainty." pith.science (2026). https://pith.science/paper/LDAI7WG6

@misc{pith2026241202886,
  author       = {Pith},
  title        = {Pith review of: Patchfinder: Leveraging Visual Language Models for Accurate Information Retrieval using Model Uncertainty},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LDAI7WG6}},
  note         = {Machine review of arXiv:2412.02886}
}
read the original abstract

For decades, corporations and governments have relied on scanned documents to record vast amounts of information. However, extracting this information is a slow and tedious process due to the sheer volume and complexity of these records. The rise of Vision Language Models (VLMs) presents a way to efficiently and accurately extract the information out of these documents. The current automated workflow often requires a two-step approach involving the extraction of information using optical character recognition software and subsequent usage of large language models for processing this information. Unfortunately, these methods encounter significant challenges when dealing with noisy scanned documents, often requiring computationally expensive language models to handle high information density effectively. In this study, we propose PatchFinder, an algorithm that builds upon VLMs to improve information extraction. First, we devise a confidence-based score, called Patch Confidence, based on the Maximum Softmax Probability of the VLMs' output to measure the model's confidence in its predictions. Using this metric, PatchFinder determines a suitable patch size, partitions the input document into overlapping patches, and generates confidence-based predictions for the target information. Our experimental results show that PatchFinder, leveraging Phi-3v, a 4.2-billion-parameter VLM, achieves an accuracy of 94% on our dataset of 190 noisy scanned documents, outperforming ChatGPT-4o by 18.5 percentage points.

Figures

Figures reproduced from arXiv: 2412.02886 by the authors.

Figure 1
Figure 1. Illustration of PatchFinder’s Confidence-Based prediction. The input image is first partitioned into multiple patches ( [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Patch Confidence as a function of patch size in the Col [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Examples of different drilling completion reports and well records. It can be seen that the data is not only highly different in [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Correlation between Accuracy and Patch Confidence [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative Examples. Green indicates a correct token, red indicates an incorrect token. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 18 canonical work pages

  1. [1]

    Strategy for digitizing archival materials for public access, 2015-2024,

    National Archives and Records Administration, “Strategy for digitizing archival materials for public access, 2015-2024,” Dec. 2014, ac- cessed: 2024-07-14. [Online]. Available: https: //www.archives.gov/digitization/strategy.html 1

  2. [2]

    DocParser: End-to-end OCR-free Information Extraction from Visually Rich Documents

    M. Dhouib, G. Bettaieb, and A. Shabou, “Docparser: End-to-end ocr-free information extraction from visually rich documents,” 2023. [Online]. Available: https://arxiv.org/abs/2304.12484 1

  3. [3]

    Pick: Processing key information extraction from documents using improved graph learning- convolutional networks,

    W. Yu, N. Lu, X. Qi, P. Gong, and R. Xiao, “Pick: Processing key information extraction from documents using improved graph learning- convolutional networks,” 2020. [Online]. Available: https://arxiv.org/abs/2004.07464 1

  4. [4]

    Ocr- free document understanding transformer,

    G. Kim, T. Hong, M. Yim, J. Nam, J. Park, J. Yim, W. Hwang, S. Yun, D. Han, and S. Park, “Ocr- free document understanding transformer,” 2022. [Online]. Available: https://arxiv.org/abs/2111.15664 1, 3

  5. [5]

    Lmdx: Language model-based document information extraction and localization,

    V . Perot, K. Kang, F. Luisier, G. Su, X. Sun, R. S. Boppana, Z. Wang, Z. Wang, J. Mu, H. Zhang, C.-Y . Lee, and N. Hua, “Lmdx: Language model-based document information extraction and localization,” 2024. [Online]. Available: https: //arxiv.org/abs/2309.10952 1

  6. [6]

    Information extraction from historical well records using a large language model,

    Z. Ma, J. E. Santos, G. Lackey, H. Viswanathan, and D. O’Malley, “Information extraction from historical well records using a large language model,” Scientific Reports, 2024, in press. 1, 3, 6

  7. [7]

    Automated structured data extraction from scanned document images,

    S. Nigam, “Automated structured data extraction from scanned document images,” inData Management, An- alytics and Innovation , S. Goswami, I. S. Barara, A. Goje, C. Mohan, and A. M. Bruckstein, Eds. Sin- gapore: Springer Nature Singapore, 2023, pp. 47–60. 1

  8. [8]

    An overview of the tesseract ocr engine,

    R. Smith, “An overview of the tesseract ocr engine,” in Ninth International Conference on Document Analysis and Recognition (ICDAR 2007), vol. 2, 2007, pp. 629–

Show all 30 references
  1. [9]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in International conference on machine learning. PMLR, 2021, pp. 8748–8763. 2, 4

  2. [10]

    Blip-2: Boot- strapping language-image pre-training with frozen im- age encoders and large language models,

    J. Li, D. Li, S. Savarese, and S. Hoi, “Blip-2: Boot- strapping language-image pre-training with frozen im- age encoders and large language models,” in Interna- tional conference on machine learning. PMLR, 2023, pp. 19 730–19 742. 2

  3. [11]

    An image is worth 16x16 words: Transformers for image recogni- tion at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weis- senborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gellyet al., “An image is worth 16x16 words: Transformers for image recogni- tion at scale,”arXiv preprint arXiv:2010.11929, 2020. 2

  4. [12]

    Vision- language models for vision tasks: A survey,

    J. Zhang, J. Huang, S. Jin, and S. Lu, “Vision- language models for vision tasks: A survey,” IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 2024. 2

  5. [13]

    Fusecap: Leveraging large language models for enriched fused image captions,

    N. Rotstein, D. Bensa ¨ıd, S. Brody, R. Ganz, and R. Kimmel, “Fusecap: Leveraging large language models for enriched fused image captions,” in Pro- ceedings of the IEEE/CVF Winter Conference on Ap- plications of Computer Vision, 2024, pp. 5689–5700. 2

  6. [14]

    Bert: Pre-training of deep bidirectional trans- formers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional trans- formers for language understanding,” arXiv preprint arXiv:1810.04805, 2018. 2

  7. [15]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Ka- plan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sas- try, A. Askell et al., “Language models are few-shot learners,” Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020. 2

  8. [16]

    Minigpt-4: Enhancing vision-language under- standing with advanced large language models,

    D. Zhu, J. Chen, X. Shen, X. Li, and M. Elho- seiny, “Minigpt-4: Enhancing vision-language under- standing with advanced large language models,”arXiv preprint arXiv:2304.10592, 2023. 2

  9. [18]

    Confidence estimation of classification based on the distribution of the neural network output layer,

    A. A. Taha, L. Hennig, and P. Knoth, “Confidence estimation of classification based on the distribution of the neural network output layer,” 2022. [Online]. Available: https://arxiv.org/abs/2210.07745 2

  10. [19]

    Consortium advancing technology for assessment of lost oil & gas wells (catalog),

    U.S. Department of Energy, “Consortium advancing technology for assessment of lost oil & gas wells (catalog),” 2024, accessed: 2024-07-15. [Online]. Available: https://catalog.energy.gov/ 2

  11. [20]

    Llava-next: Improved reasoning, ocr, and world knowledge,

    H. Liu, C. Li, Y . Li, B. Li, Y . Zhang, S. Shen, and Y . J. Lee, “Llava-next: Improved reasoning, ocr, and world knowledge,” January 2024. [Online]. Available: https://llava-vl.github.io/blog/2024-01-30-llava-next/ 3

  12. [21]

    Phi-3 technical report: A highly capable language model locally on your phone,

    M. Abdin, S. A. Jacobs, A. A. Awan, J. Aneja, A. Awadallah, H. Awadalla, N. Bach, A. Bahree, A. Bakhtiari, J. Bao, H. Behl, A. Benhaim, M. Bilenko, J. Bjorck, S. Bubeck, Q. Cai, M. Cai, C. C. T. Mendes, W. Chen, V . Chaudhary, D. Chen, D. Chen, Y .-C. Chen, Y .-L. Chen, P. Cho...

  13. [22]

    Idefics2: A powerful 8b vision- language model for the community,

    HuggingFaceM4, “Idefics2: A powerful 8b vision- language model for the community,” Hugging Face,

  14. [23]

    Improved baselines with visual instruction tuning,

    H. Liu, C. Li, Y . Li, and Y . J. Lee, “Improved baselines with visual instruction tuning,” 2024. [Online]. Available: https://arxiv.org/abs/2310.03744 3

  15. [24]

    Cloud vision api,

    Google, “Cloud vision api,” https://cloud.google.com/ vision, accessed: 2024-07-15. 3

  16. [25]

    A baseline for detect- ing misclassified and out-of-distribution examples in neural networks,

    D. Hendrycks and K. Gimpel, “A baseline for detect- ing misclassified and out-of-distribution examples in neural networks,” arXiv preprint arXiv:1610.02136 ,

  17. [26]

    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. 4

  18. [27]

    Llama: Open and ef- ficient foundation language models,

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozi `ere, N. Goyal, E. Hambro, F. Azhar et al. , “Llama: Open and ef- ficient foundation language models,” arXiv preprint arXiv:2302.13971, 2023. 5

  19. [28]

    Unlocking solutions: In- novative approaches to identifying and mitigating the environmental impacts of undocumented orphan wells in the united states,

    D. O’Malley, A. A. Delorey, E. J. Guiltinan, Z. Ma, T. Kadeethum, G. Lackey, J. Lee, J. E. Santos, E. Fol- lansbee, M. C. Nair et al., “Unlocking solutions: In- novative approaches to identifying and mitigating the environmental impacts of undocumented orphan wells in the unit...

  20. [29]

    Chatgpt-4o,

    OpenAI, “Chatgpt-4o,” https : / / www. openai . com, 2024, large language model trained by OpenAI. 7

  21. [30]

    Corporate reports online,

    Penn Libraries, “Corporate reports online,” https://colenda.library.upenn.edu/?f7

  22. [2024]

    Available: https://huggingface.co/ HuggingFaceM4/idefics2-8b 3

    [Online]. Available: https://huggingface.co/ HuggingFaceM4/idefics2-8b 3

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.