Pith. sign in

REVIEW 4 major objections 5 minor 25 references

Mind the Gap: Entity-Preserved Context-Aware ASR Structured Transcriptions

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Giving a speech recognizer five extra seconds of audio on each side of its training chunk, while evaluating only the central part, improves how it recognizes and formats names and numbers.

desk verdict A practical windowing idea for ASR entity formatting, but the main comparison is confounded and the inference procedure drops the right context used in training. read the letter →

arxiv 2506.22858 v1 pith:KLEY6SED submitted 2025-06-28 cs.CL cs.SDeess.AS

Duygu Altinok This is my paper · ORCID
classification cs.CLcs.SDeess.AS
keywords automaticspeechrecognitionnamedentityformattingoverlappingcontextwindowsWhisperfine-tuninglong-formtranscriptionspokenWikipediasequencelabeling
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 claims that a speech recognizer can be made better at transcribing named entities and numbers by training it on audio chunks that include five extra seconds of context on each side. The proposal, applied to the Whisper-Medium model, uses a 40-second input window while computing the prediction loss only on the central 30 seconds, so inference does not see longer chunks than usual. Entities that would be split by a window boundary are moved entirely into the right-hand chunk, and entity-type tags are embedded in the training text so the model learns both recognition and type-specific formatting. On the Spoken Wikipedia corpus, the windowed model improves named-entity F1 and numeric formatting over the same model trained on plain 30-second chunks. If the effect holds, long-form ASR and spoken-language understanding can gain semantic accuracy without a more expensive inference pipeline.

What carries the argument

The load-bearing mechanism is the 'effective semantic window': a 40-second training clip formed by sliding 5-second overlaps onto both sides of a 30-second target chunk. The model's encoder is extended with new positional-embedding rows, initialised to match the distribution of the original embeddings and warmed up on 100 hours of concatenated VoxPopuli audio, and the decoder is trained with a masked cross-entropy loss that counts only tokens in the central 30-second span, signalled by the special tokens <|left|>, <|mid|>, and <|right|>. Entities crossing a boundary are reassigned to the right-hand chunk, which teaches the model to defer a prediction until the full entity is visible. This combination allows longer context to inform each prediction without changing the inference-time chunking.

What would settle it

Train a control model on the same 40-second inputs and masked central-window loss, but with the left and right 5-second overlap regions replaced by silence; if its entity F1 and numeric character error rate match Whisper-puncted-windowed, the reported gains come from longer input or the embedding warm-up, not from surrounding semantic context.

Watch

Extended reading notes

Core claim

The central discovery is that a fixed-length ASR model can be taught to use a wider semantic context at no inference cost by decoupling the input window from the prediction window. Whisper-puncted-windowed is trained on 40-second clips (left 5 seconds, middle 30 seconds, and right 5 seconds), with cross-entropy loss masked so only tokens between <|mid|> and <|right|> count, and with boundary-spanning entities shifted to the right-hand chunk. Compared with Whisper-puncted, a Whisper-Medium model fine-tuned on the same entity-tagged transcripts in plain 30-second chunks, this model reports higher named-entity recognition F1 for most entity types (e.g., PERSON 0.50 to 0.65, GPE 0.61 to 0.71, CARDINAL 0.95 to 0.98) and lower character error rates on numeric entities (e.g., MONEY 0.42 to 0.11, PERCENT 0.32 to 0.12, PHONE_NUM 0.35 to 0.15). The paper interprets these gains as evidence that the 10 extra seconds of overlapping context let the decoder recognize entity boundaries and formats that are ambiguous in isolation.

Load-bearing premise

The claim depends on the assumption that the extra positional-embedding rows, warmed up on only 100 hours of concatenated audio, genuinely enable the model to exploit the 10 extra seconds of audio rather than merely pass them through unprocessed.

Editorial extensions

If this is right

  • If the reported gains are real, the same model can be deployed on long-form audio with its standard strided inference, so better entity recognition does not require a larger or slower recognizer.
  • The entity-reassignment rule should remove the characteristic error in which a monetary or numeric entity is split across chunks and mis-tagged as two entities.
  • Masking the loss to the central window turns the overlap region into free context for every training example, which could be applied to other sequence-to-sequence ASR models with fixed positional embeddings.
  • Embedding entity tags in the training transcript teaches the decoder to emit type-specific punctuation and casing, giving downstream text-based NER a cleaner input.

Reading between the lines

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

  • A testable extension would be to apply the same sliding-window masking to entity-aware speech translation, where the encoder context may benefit from future audio before a spoken entity is complete.
  • The paper leaves open whether the gains come from the overlapping context itself or from the extra 100-hour embedding warm-up; an ablation with the same 40-second inputs and the overlap regions replaced by silence would separate the two.
  • The boundary-reassignment rule could be generalised to any multi-token unit, such as dates or other numerical expressions, wherever the decoder needs to see the end of a unit before committing.
  • Because the method changes only the training data layout and loss mask, it could be combined with decoder prompting or language-model rescoring to further reduce residual punctuation errors inside entities.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes an ASR fine-tuning approach for entity-aware transcription: Whisper-puncted-windowed is trained on 40-second chunks (5s left, 30s middle, 5s right), with a masked loss over the central 30 seconds, NER tags embedded in transcripts, and boundary-crossing entities shifted entirely into the right-hand chunk. A baseline Whisper-puncted model is fine-tuned on 30-second tagged chunks. On the Spoken Wikipedia dataset, the windowed model reports higher seqeval F1 for most entity types and lower CER for numerical entities, and the authors conclude that extended semantic windows improve ASR performance. The paper also introduces entity-formatting metrics for numerical and textual entities and releases code and models.

Significance. If the attribution to the extended context window were cleanly established, this would be a useful contribution: the boundary-shift idea is practical and clearly described, the per-entity-type evaluation with separate CER and Jaro-Winkler metrics is informative, and the public release of code and fine-tuned models supports reproducibility. The central comparison, however, is currently confounded by several simultaneous changes between baseline and proposed model, and the inference procedure does not match the training setup. As a result, the headline quantitative claims are not yet supported, although the underlying approach is worth investigating further with additional ablations and reporting.

major comments (4)
  1. [Section 4.5 vs. Sections 4.1 and 4.4] Section 4.5 states that during inference the model processes 35-second left+mid audio segments (30 seconds for the first segment), while Sections 4.1 and 4.4 train on 40-second left+mid+right chunks with loss masked to the mid window. The deployed setup therefore removes the right (future) 5 seconds of audio that the model saw at training time, so the effective semantic window at inference is 35 seconds, not the 40 seconds claimed in the abstract. Moreover, training shifts entities crossing the mid-right boundary entirely into the right window (Figure 1), teaching the model to delay their emission; at inference that right audio is absent, which can cause those entities to be dropped or misaligned. Because Section 6 does not report WER for Whisper-puncted-windowed, the reader cannot tell whether this mismatch degrades overall transcription. The paper should either evaluate with a 40-second input that provides the right context, or add an ablation trained and evaluated with the 35-second left+mid input used at inference, and should report WER for the proposed model.
  2. [Section 4.3 and Section 5] The comparison between Whisper-puncted and Whisper-puncted-windowed is confounded: the windowed model differs from the baseline not only in the 10 extra seconds of audio, but also in newly added positional embeddings pretrained on 100 hours of concatenated VoxPopuli audio with inserted silences, in three extra special tokens (<|left|>, <|mid|>, <|right|>), and in a different prompting structure. The models are also trained for different numbers of epochs (100 versus 150) with checkpoint selection reported at different epochs. Without ablations that isolate the context window from these other changes, the reported gains cannot be attributed specifically to the extended semantic window, and the introduction's claim that the results underscore the value of extended semantic windows is stronger than the experiments support.
  3. [Section 6.1 and Tables 1-2] Section 6.1 reports WER only for the unmodified Whisper-Medium model (38%) and for Whisper-puncted (26%); no WER is reported for Whisper-puncted-windowed. Since NER F1 and formatting CER are computed on the model's own transcripts, a change in overall transcription quality could inflate or deflate the semantic metrics independently of the windowing mechanism. Reporting WER for the proposed model, and ideally entity-conditioned WER, is necessary to interpret the improvements in Tables 1 and 2.
  4. [Tables 1 and 2 and Abstract/Introduction] The claimed improvements "across all metrics" are not supported by the tables. In Table 2, Jaro-Winkler scores are identical for PERSON (0.75/0.75), ORG (0.71/0.71), EVENT, LAW, FAC, PROD, LANG, and W_OF_ART, and in Table 1 DATE F1 is unchanged at 0.65. Section 6.2 itself acknowledges only "marginal gains" for textual entities. The abstract and introduction should be revised to claim improvements only for the specific entity types and metrics where they are observed, and the paper should include significance tests or confidence intervals, especially for low-count types such as MONEY (109 instances) and QUANTITY (351 instances).
minor comments (5)
  1. [Section 4.3] The text refers to "Whisper-puncted-segmented" but the model under discussion is called "Whisper-puncted-windowed" elsewhere; the naming should be made consistent.
  2. [Section 4.4] Equation (7) defines the mask using tmid and tright but never defines tleft; the paper should define all boundary positions used in the mask and clarify whether the mask includes the special boundary tokens themselves.
  3. [Section 4.5] The description of inference for the first segment (30 seconds with no left context) and for subsequent segments (35 seconds with left context) does not state how these two conditions are pooled in the reported metrics; this should be clarified.
  4. [Section 2] The citation "Jannet et al., 2017" in the related work should be "Ben Jannet et al., 2017" to match the reference list entry.
  5. [Tables 1 and 2] The column headers "W-punct" and "W-pwindow" are not expanded anywhere; adding a note in the table captions would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivational circularity: the paper's claims are empirical comparisons against an external benchmark, with standard masked cross-entropy training and no load-bearing self-citations.

full rationale

The paper does not derive any result from its own definitions in a circular way. Its central comparison is empirical: Whisper-puncted-windowed is fine-tuned on 40-second chunks with a masked loss over the central 30 seconds and evaluated on the held-out Spoken Wikipedia test set against a similarly fine-tuned 30-second baseline. The loss in Equations 5-8 is standard cross-entropy with label smoothing and a masking term, and the mask is defined directly in terms of token positions relative to <|mid|> and <|right|>, not in terms of the evaluation metrics. The entity-boundary reassignment is a deterministic data-preparation rule, not a fitted parameter later renamed as a prediction. The paper contains no self-citations at all, and no uniqueness claim is imported from the authors' prior work. The closest concern is the train/inference mismatch described in Section 4.5: training uses 40-second left+mid+right inputs while inference uses 35-second left+mid inputs. This is a possible confound for attributing the gains to 'extended semantic windows,' but it is an experimental-design and correctness concern, not circular reasoning. No equation reduces to an input, no fitted quantity is presented as a prediction, and no internally justified premise is used to force the conclusion. Therefore the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The paper's central claim rests on empirical training plus assumptions about Whisper transfer and dataset label quality. The main design parameters, window sizes and training durations, are chosen by hand, and the evaluation depends on the accuracy of spaCy and regex-based annotations.

free parameters (3)
  • Context window sizes = 5 s overlap, 30 s central, 40 s effective input
    Hand-chosen design parameters; no ablation varying window lengths is reported despite the abstract claiming such analysis. The sizes define the proposed method and the reported gains.
  • Label smoothing factor epsilon = not reported
    Required in Equation 5 but no value is given in the experimental setup; affects training loss but is a routine hyperparameter.
  • Training epochs and checkpoint selection = baseline: 100 epochs, best at 80; windowed: 150 epochs
    Chosen by monitoring during training; no validation split is documented, so checkpoint selection may have been influenced by the test set.
assumptions (4)
  • domain assumption Whisper's pretrained weights remain usable after extending encoder positional embeddings with randomly initialized rows and light VoxPopuli training.
    Invoked in Section 4.3 when extending input from 30 seconds to 40 seconds; no ablation or fine-tuning analysis confirms this.
  • domain assumption Gold word timestamps in Spoken Wikipedia are accurate enough to chunk audio and to shift boundary entities without corrupting audio-text alignment.
    Chunking relies on token-level timestamps (Sections 3 and 4.1); if timestamps are noisy, the right-hand entity reassignment may misalign audio and text.
  • domain assumption spaCy NER and the custom regex extractor produce correct entity labels for both training and test data.
    Used to create all entity tags (Section 3); errors in these labels would contaminate both training targets and evaluation.
  • domain assumption Label smoothing and masking only the central window leave the pretrained decoder's sequence-level behavior intact.
    The loss in Section 4.4 restricts training to the middle window, but the paper does not analyze whether this disrupts Whisper's learned decoding patterns.
invented entities (1)
  • Special window tokens <|left|>, <|mid|>, <|right|>
    purpose: Mark window boundaries in the target transcript and instruct the model to generate text only in the central window.
    Introduced in Section 4.2 and used to mask loss and prompt inference; no external evidence beyond this paper's experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mind the Gap: Entity-Preserved Context-Aware ASR Structured Transcriptions." pith.science (2026). https://pith.science/paper/KLEY6SED

@misc{pith2026250622858,
  author       = {Pith},
  title        = {Pith review of: Mind the Gap: Entity-Preserved Context-Aware ASR Structured Transcriptions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KLEY6SED}},
  note         = {Machine review of arXiv:2506.22858}
}
read the original abstract

Automatic Speech Recognition (ASR) systems, such as Whisper, achieve high transcription accuracy but struggle with named entities and numerical data, especially when proper formatting is required. These issues increase word error rate (WER) and impair semantic understanding in critical domains like legal, financial, and medical applications. We propose a novel training approach that extends the semantic context of ASR models by adding overlapping context windows during training. By sliding 5-second overlaps on both sides of 30-second chunks, we create a 40-second "effective semantic window," improving entity recognition and formatting while focusing predictions on the central 30 seconds. To address entities spanning chunk boundaries, we reassign such entities entirely to the right-hand chunk, ensuring proper formatting. Additionally, enriched training data with embedded entity labels enables the model to learn both recognition and type-specific formatting. Evaluated on the Spoken Wikipedia dataset, our method improves performance across semantic tasks, including named entity recognition (NER) and entity formatting. These results highlight the effectiveness of context-aware training in addressing ASR limitations for long-form transcription and complex entity recognition tasks.

Figures

Figures reproduced from arXiv: 2506.22858 by the authors.

Figure 1
Figure 1. Examples illustrating our windowing methodology. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. An example from the dataset, where each word is accompanied by its pronunciation, start time, and end time. This detailed alignment facilitates effi￾cient chunking of the audio. Additionally, multi-word entities are annotated with word-level information for each constituent word [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Distribution of NER labels in the dataset. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Two examples of reverse normalization. The first example illustrates a reverse-normalized chunk text for [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Special token placement for a standard chunk and a windowed chunk.. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: An example of a numerical entity transcription error caused by spreading across chunk boundaries. The [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 12 canonical work pages

  1. [1]

    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 INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Rosana Ardila, Megan Branson, Kelly Davis, Michael Kohler, Josh Meyer, Michael Henretty, Reuben Morais, Lindsay Saunders, Francis Tyers, and Gregor Weber. 2020. https://aclanthology.org/2020.lrec-1.520/ Common voice: A massively-multilingual speech corpus . In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 4218--4222, Marse...

    work page 2020

  4. [4]

    Siddhant Arora, Siddharth Dalmia, Brian Yan, Florian Metze, Alan W Black, and Shinji Watanabe. 2022. https://doi.org/10.18653/v1/2022.findings-emnlp.396 Token-level sequence labeling for spoken language understanding using compositional end-to-end models . In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 5419--5429, Abu Dhab...

    work page doi:10.18653/v1/2022.findings-emnlp.396 2022

  5. [5]

    Timo Baumann, Arne K \"o hn, and Felix Hennig. 2018. https://doi.org/10.1007/s10579-017-9410-y The spoken wikipedia corpus collection: Harvesting, alignment and an application to hyperlistening . Language Resources and Evaluation

    doi:10.1007/s10579-017-9410-y 2018

  6. [6]

    Boli Chen, Guangwei Xu, Xiaobin Wang, Pengjun Xie, Meishan Zhang, and Fei Huang. 2022. https://doi.org/10.1109/ICASSP43922.2022.9746955 Aishell-ner: Named entity recognition from chinese speech . In ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 8352--8356

    arXiv 2022

  7. [7]

    Marco Gaido, Stefano Papi, Matteo Negri, and Marco Turchi. 2023. https://doi.org/10.21437/Interspeech.2023-1767 Joint speech translation and named entity recognition . In Proceedings of Interspeech 2023, pages 47--51

    work page doi:10.21437/interspeech.2023-1767 2023

  8. [8]

    Matthew Honnibal, Ines Montani, Sofie Van Landeghem, and Adriane Boyd. 2020. https://doi.org/10.5281/zenodo.1212303 spacy 2: Natural language understanding with bloom embeddings, convolutional neural networks and incremental parsing

    doi:10.5281/zenodo.1212303 2020

Show all 25 references
  1. [9]

    Mohamed Ameur Ben Jannet, Olivier Galibert, Martine Adda-Decker, and Sophie Rosset. 2017. https://doi.org/10.21437/Interspeech.2017-1482 Investigating the effect of asr tuning on named entity recognition . In Interspeech 2017, pages 2486--2490

    2017 doi

  2. [10]

    Matthew A. Jaro. 1989. Advances in record-linkage methodology as applied to the 1985 census of tampa florida. Journal of the American Statistical Association, 84(406):414--420

    1989

  3. [11]

    Yifan Li, Jun Yu, Ming Zhang, Min Ren, Yan Zhao, Xiang Zhao, Shiyi Tao, Jie Su, and Han Yang. 2024. https://doi.org/10.21437/Interspeech.2024-103 Using large language model for end-to-end chinese asr and ner . In Proceedings of Interspeech 2024, pages 822--826

    2024 doi

  4. [12]

    Ilya Loshchilov and Frank Hutter. 2019. https://openreview.net/forum?id=Bkg6RiCqY7 Decoupled weight decay regularization . In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019 . OpenReview.net

    2019

  5. [13]

    Salima Mdhaffar, Jarod Duret, Titouan Parcollet, and Yannick Estève. 2022. https://doi.org/10.21437/Interspeech.2022-10231 End-to-end model for named entity recognition from speech without paired training data . In Interspeech 2022, pages 4068--4072

    2022 doi

  6. [14]

    Hiroki Nakayama. 2018. seqeval: A python framework for sequence labeling evaluation. https://github.com/chakki-works/seqeval

    2018

  7. [15]

    Vassil Panayotov, Guoguo Chen, Daniel Povey, and Sanjeev Khudanpur. 2015. https://doi.org/10.1109/ICASSP.2015.7178964 Librispeech: An asr corpus based on public domain audio books . In 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), page...

    2015

  8. [16]

    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. https://arxiv.org/abs/2212.04356 Robust speech recognition via large-scale weak supervision . Preprint, arXiv:2212.04356

    2022 arXiv

  9. [17]

    Anthony Rousseau, Paul Del \'e glise, and Yannick Est \`e ve. 2012. https://aclanthology.org/L12-1405/ TED - LIUM : an automatic speech recognition dedicated corpus . In Proceedings of the Eighth International Conference on Language Resources and Evaluation ( LREC `12) , pages...

    2012

  10. [18]

    Dmitriy Serdyuk, Yongqiang Wang, Christian Fuegen, Anuj Kumar, Baiyang Liu, and Yoshua Bengio. 2018. https://doi.org/10.1109/ICASSP.2018.8461785 Towards end-to-end spoken language understanding . In 2018 IEEE International Conference on Acoustics, Speech and Signal Processing ...

    2018

  11. [19]

    Piotr Szyma \'n ski, Lukasz Augustyniak, Mikolaj Morzy, Adrian Szymczak, Krzysztof Surdyk, and Piotr \.Z elasko. 2023. https://doi.org/10.18653/v1/2023.acl-long.98 Why aren`t we NER yet? artifacts of ASR errors in named entity recognition in spontaneous speech transcripts . In...

    2023 doi

  12. [20]

    Changhan Wang, Morgane Riviere, Ann Lee, Anne Wu, Chaitanya Talnikar, Daniel Haziza, Mary Williamson, Juan Pino, and Emmanuel Dupoux. 2021. https://aclanthology.org/2021.acl-long.80 V ox P opuli: A large-scale multilingual speech corpus for representation learning, semi-superv...

    2021

  13. [21]

    William E. Winkler. 1990. String comparator metrics and enhanced decision rules \\ in the fellegi-sunter model of record linkage. Technical report, US Census Bureau

    1990

  14. [22]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mari...

    2020 doi

  15. [23]

    Hardik Yadav, Soujanya Ghosh, Yong Yu, and Rajiv Ratn Shah. 2020. https://doi.org/10.21437/Interspeech.2020-2482 End-to-end named entity recognition from english speech . In Proceedings of Interspeech 2020, pages 4268--4272

    2020 doi

  16. [24]

    , " * write output.state after.block = add.period write

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution journal key month note number organization pages publisher school series title type url volume year label INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION in...

  17. [25]

    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 6, 2026 · model on record in the stance chip above.