A surprisal oracle for when every layer counts

Asad Sayeed; Sharid Lo\'aiciga; Xudong Hong

REVIEW 4 major objections 3 minor 17 references

A surprisal oracle for when every layer counts

T0 review · 4 major / 3 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a dynamically updated surprisal curriculum, driven by the model's own uncertainty, improves a strong small language model on world-knowledge and common-sense tasks over official baselines.

desk verdict Modest, honest BabyLM system paper whose headline EWOK/GLUE gains are real but not yet attributable to ACLM because the controls are confounded on effective batch size. read the letter →

arxiv 2412.03098 v1 pith:7CJFFIR7 submitted 2024-12-04 cs.CL

Xudong Hong , Sharid Loáiciga , Asad Sayeed This is my paper · ORCID

classification cs.CL

keywords activecurriculumlearningsurprisaloracleBabyLMELC-BERTsample-efficientlanguagemodelinglearner-directedworldknowledgeevaluationdynamic

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 argues that letting a language model steer its own training order, by repeatedly re-measuring which sentences in its pool are most surprising to it at that moment, can push a strong small architecture to better world-knowledge and common-sense scores. This is the 'dynamic surprisal oracle': each iteration the model picks its least-certain training item, finds pool sentences with similar surprisal profiles, adds them to the active set, and rebuilds the surprisal space from its current state. The authors apply this procedure on top of ELC-BERT and report consistent gains over the official BabyLM 2024 baselines (LTG-BERT and BabyLlama) on EWOK and GLUE, alongside weaker grammatical-inference scores on BLiMP. A reader should care because the result suggests that, independent of architecture and data volume, the order in which a learner encounters data is a handle worth controlling. The authors are careful to note that without significance testing the score gaps are provisional.

What carries the argument

The mechanism is the active curriculum itself: at each iteration, the model under training (ELC-BERT) computes per-sentence average surprisal over its active training set, selects the sentence with the highest surprisal as the 'least certain' item, and then uses $k$-nearest-neighbours on the surprisal vectors of the as-yet-unseen pool to pull in the most similar items, which join the training set for the next epoch. The update from the previous year's ACLM is that the surprisal space is re-computed every iteration from the current model rather than fixed in advance from a trigram model, making the curriculum dynamically learner-directed. The surprisal vectors are reduced to 7, 32, 64, or 128 dimensions, and the paper finds that this dimensionality choice makes little difference once the input length is uniformly 128 tokens.

What would settle it

Re-run the ACLM and non-ACLM ELC-BERT models under identical settings for, say, ten random seeds, compute confidence intervals on EWOK and GLUE accuracies, and check whether the ACLM advantage over its own control persists. If the gap shrinks to within the noise, the central claim that the dynamic surprisal curriculum drives the improvement collapses; likewise, if a control with randomly selected batches under the same update schedule matches ACLM, then the specific surprisal-based selection is not the cause.

Watch

Extended reading notes

Core claim

The central claim is that a learner-directed curriculum, updated by the model's own changing notion of surprisal, improves a strong data-constrained language model on world-knowledge and common-sense tasks relative to passive training. Concretely, ACLM over ELC-BERT scores 70.0–72.1 on EWOK against 48.9 (LTG-BERT) and 50.7 (BabyLlama), and 63.5–65.7 on GLUE against 60.3 and 63.3, while the non-ACLM ELC-BERT runs sit at 64.9–65.2 on EWOK and 61.0–63.4 on GLUE. The authors interpret this as evidence that controlling the order of learning can matter, and they connect it speculatively to developmental dependencies among tasks; they also report that the same models underperform on fine-grained grammatical inferences (BLiMP), which they attribute mostly to the much smaller batch size they used in all their runs. They do not claim to have a complete account of why the curriculum helps, and they explicitly flag the absence of significance tests, multiple-seed averages, and an exploration of hyperparameter space as limitations.

Load-bearing premise

The load-bearing premise is that the reported score gaps between ACLM and its non-ACLM control reflect the curriculum rather than run-to-run noise or evaluation details, since the paper reports no significance tests, confidence intervals, or multiple-seed averages.

Editorial extensions

If this is right

Learning order is a real lever: the same architecture, data, and batch size produce better EWOK and GLUE scores when the training set is curated by the model's own surprisal rather than fixed in advance.
The curriculum can be a portable add-on: ACLM is model-agnostic in principle, and here it transfers to a different base architecture (ELC-BERT) from last year's RoBERTa-based system.
Hyperparameter sensitivity of strong baselines can confound curriculum comparisons: the non-ACLM ELC-BERT runs underperform on BLiMP, which the authors attribute to batch size, so future comparisons need matched hyperparameters to isolate the curriculum effect.
The dimensionality of the surprisal signature may not matter when input lengths are uniform, so future work should test variable-length inputs where the reduction scheme could bite.

Reading between the lines

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

A testable extension the paper does not run: apply ACLM on top of the original full-size ELC-BERT configuration rather than the batch-size-limited runs, to see whether the curriculum still improves EWOK and GLUE when the base model is not handicapped.
The paper's cognitive reading suggests a direct probe: inspect the actual order of sentences selected by ACLM and test whether it correlates with human developmental orderings (e.g., concreteness or syntactic complexity), which the paper only speculates about.
If the EWOK and GLUE gains survive significance testing, the surprisal criterion may be capturing a general 'zone of proximal development' signal that could be combined with other informativeness measures, since the authors note that a single measure is unlikely to be optimal.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 genuinely new thing here is small but real: the authors update their ACLM procedure so the surprisal space is recomputed from the model itself each iteration, and they run it over ELC-BERT for BabyLM 2024. The EWOK and GLUE numbers beat the official baselines and their own non-ACLM ELC-BERT runs, and the effect is consistent across all four surprisal-space dimensionalities. That is a reasonable empirical contribution to the BabyLM line, and the paper is written honestly—it flags the lack of significance testing, the English-only scope, and the batch-size sensitivity of BLiMP. Credit is also due for releasing code. The soft spot is not that they lack significance tests; for a workshop shared task that is a common and acknowledged limitation. The deeper problem is that the ACLM runs and the non-ACLM controls are not matched on compute. Table 2 reports ACLM with batch size 512 and gradient accumulation 4, so effective batch 2048, with 31,250 training steps. The ELC-BERT B512 row is presented as the 'very similar settings' control, but the paper never tells us its accumulation value or total steps. If B512 uses accumulation 1 (effective batch 512), or if Algorithm 2's multiple epochs per iteration push ACLM past 31,250 effective steps, then the EWOK and GLUE gaps could simply be a batch-size or compute effect. The authors themselves show that batch size matters a lot for BLiMP, so it is not safe to assume it does not matter for EWOK and GLUE. To claim the curriculum drives the improvement, they need a random-order control trained with the same effective batch, step count, and learning-rate schedule. That said, the paper does not hide these weaknesses. The Limitations section is candid, and the discussion of batch-size sensitivity is fair. The central claim—ACLM improves EWOK/GLUE—is plausible but unproven as stated. Given the shared-task context, I would not desk-reject it, but a serious referee should ask for matched-compute controls and at least a few seeds before the comparison is treated as evidence about curriculum learning. This paper will mainly interest people working on curriculum learning or on squeezing performance from small models. I would not cite it for the causal claim, but I might cite it as an example of a dynamic surprisal-based curriculum. Worth a reading-group slot for the methodology, not for the headline.

Referee Report

4 major / 3 minor

Summary. The paper presents an updated Active Curriculum Language Modeling (ACLM) system for the BabyLM 2024 strict-small track. The method iteratively selects training items using the model's own surprisal values, with a dynamically recomputed surprisal space, and is applied to ELC-BERT. The authors report results on BLiMP, EWOK, and GLUE (Table 1), showing that ACLM variants outperform both the official BabyLM 2024 baselines and their non-ACLM ELC-BERT runs on EWOK and GLUE, while underperforming on BLiMP. They also discuss batch-size sensitivity and mention several limitations, including the lack of significance testing.

Significance. If the causal claim holds, the paper provides evidence that a learner-directed, dynamically updated curriculum can improve world-knowledge and common-sense benchmarks under data-constrained conditions, which is directly relevant to the BabyLM line of work and to curriculum learning more broadly. The manuscript is transparent about its limitations and makes code publicly available, which supports reproducibility. However, the central comparative claim currently rests on single runs without significance testing and without fully matched training schedules, so the significance is conditional on additional experimental support.

major comments (4)

[§4, Table 1; Limitations] The central claim that ACLM improves over the non-ACLM ELC-BERT runs and over the official baselines on EWOK and GLUE is not supported by significance testing or multiple random seeds. The Limitations section explicitly states that 'We do not have a full analysis of the statistical significances of the differences in the scores,' and Table 1 reports single-run accuracy values. Observed gaps such as EWOK 70.0–72.1 versus 64.9–65.2 and GLUE 63.5–65.7 versus 61.0–63.4 may be within run-to-run variance; the authors should report confidence intervals, bootstrap estimates, or results over several seeds before attributing these differences to the ACLM curriculum.
[§4, Table 2, Algorithm 2] The non-ACLM ELC-BERT B32 and B512 runs are not documented as controls in sufficient detail. Table 2 lists hyperparameters for the ACLM models only, including batch size 512, gradient accumulation 4, and 31,250 training steps; the gradient-accumulation values, effective batch sizes, total optimizer steps, and number of epochs for the B32 and B512 runs are not reported. Since Algorithm 2 trains for n_epochs per iteration and the value of n_epochs is not given, the ACLM runs may have a different effective batch size (512×4 = 2048) or a different total number of updates than the B512 control. The paper should provide a matched-compute or matched-effective-batch random-order control to support the claim that the improvements are due to the curriculum rather than to schedule or compute differences.
[§4, BLiMP paragraph] The text states that 'the overall consistency of outperformance of ACLM on the filtered BLiMP suggests that ACLM is having an effect,' but Table 1 shows differences of only about one or two points (e.g., filtered BLiMP ACLM-D7 51.3 versus B512 49.1 and B32 47.9) with no significance testing and no seed variation. This statement overstates the evidence and should be removed or supported with statistical analysis.
[Abstract and §4] The abstract's claim that 'they outperform the BabyLM 2024 official baselines on common-sense and world-knowledge tasks' is accurate only for the specific single runs reported in Table 1; as written, it can be read as a general claim. The abstract should qualify the result as applying to the runs in this study, pending the significance and control analyses described above.

minor comments (3)

[Abstract, Section 2, Figure 1 caption] There are several typos and inconsistent spellings: 'surpisal' in the Figure 1 caption, 'suprisal' alongside 'surprisal' in Section 2, 'peturbations' in the Introduction, and 'acquistion' in the Introduction. These should be corrected.
[Table 2 heading] Table 2 is labeled 'Small (Submitted Model)' but it is unclear whether the same hyperparameters apply to all four ACLM variants (D7, D32, D64, D128) or whether the surprisal-space dimensionality affects any listed hyperparameters. The table caption and text should clarify this.
[Section 2, footnote 2] The parenthetical remark about a reviewer suggesting a switch in averaging procedure is informal for a journal article; consider moving such discussion to a more explicit comparison of evaluation pipelines or removing it.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the ACLM comparison is assessed on external benchmarks, and the model's own surprisal signal is the method's mechanism, not a fitted prediction.

full rationale

The paper's central claim is that ACLM-trained ELC-BERT models outperform official BabyLM 2024 baselines and non-ACLM ELC-BERT runs on EWOK and GLUE. The evidence is empirical and external: scores on EWOK, GLUE, and BLiMP are held-out benchmark results, not outputs of the training objective or fitted parameters. The surprisal signal is computed by the model under training and used to select the next training items; this is the method's intended mechanism, not a post-hoc fit to the evaluation data. No hyperparameter is selected based on benchmark performance; the reported D7/D32/D64/D128 variants are all presented without choosing a best one. The self-citation to Hong et al. (2023) describes the prior ACLM formulation and previous competition results; it is background, not a load-bearing theorem or uniqueness claim. The acknowledged limitations, including lack of significance testing and unmatched batch-size controls, weaken causal attribution but do not make the derivation circular. Accordingly, no circular step can be identified and the score is 0.

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

The central empirical claim depends on unspecified ACLM hyperparameters (initial set size, k, iterations) and on the assumption that surprisal is a good learning signal. No entities are invented, and no parameters are fitted to the target benchmarks.

free parameters (4)

Initial active set size (n_initial)
Algorithm 1 samples n_initial sentences from the pool; the value is not reported, yet it determines the starting curriculum.
k-nearest neighbours (k)
Algorithm 2 adds k nearest surprisal neighbors per iteration; k is not reported.
Number of ACLM iterations (n_iterations)
The number of curriculum update cycles is not stated.
Surprisal space dimensionality = 7, 32, 64, 128
Hand-selected dimensions for dimensionality reduction; results are similar across settings, so it does not drive the main claim.

assumptions (4)

domain assumption The model's own surprisal is a valid uncertainty signal for selecting training items.
Section 3 describes using the model's average sentence surprisal to identify the least certain item and retrieve similar ones.
domain assumption Curriculum learning can improve language model performance under data constraints.
Motivates the approach; references Bengio et al. (2009) and active learning literature.
domain assumption ELC-BERT is a suitable base model for the strict-small task.
Chosen because it was the most successful BabyLM 2023 model; the authors replicate it with smaller batches.
domain assumption The BabyLM evaluation tasks measure relevant language abilities.
The paper uses BLiMP, EWOK, and GLUE as the official evaluation suite.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A surprisal oracle for when every layer counts." pith.science (2026). https://pith.science/paper/7CJFFIR7

@misc{pith2026241203098,
  author       = {Pith},
  title        = {Pith review of: A surprisal oracle for when every layer counts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7CJFFIR7}},
  note         = {Machine review of arXiv:2412.03098}
}

read the original abstract

Active Curriculum Language Modeling (ACLM; Hong et al., 2023) is a learner directed approach to training a language model. We proposed the original version of this process in our submission to the BabyLM 2023 task, and now we propose an updated ACLM process for the BabyLM 2024 task. ACLM involves an iteratively- and dynamically-constructed curriculum informed over the training process by a model of uncertainty; other training items that are similarly uncertain to a least certain candidate item are prioritized. Our new process improves the similarity model so that it is more dynamic, and we run ACLM over the most successful model from the BabyLM 2023 task: ELC-BERT (Charpentier and Samuel, 2023). We find that while our models underperform on fine-grained grammatical inferences, they outperform the BabyLM 2024 official base-lines on common-sense and world-knowledge tasks. We make our code available at https: //github.com/asayeed/ActiveBaby.

Figures

Figures reproduced from arXiv: 2412.03098 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 5 canonical work pages

[1]

URL: " 'urlintro :=

ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...
[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]

Yoshua Bengio, J\' e r\^ o me Louradour, Ronan Collobert, and Jason Weston. 2009. https://doi.org/10.1145/1553374.1553380 Curriculum learning . In Proceedings of the 26th Annual International Conference on Machine Learning, ICML '09, page 41–48, New York, NY, USA. Association for Computing Machinery

arXiv 2009
[4]

Lucas Georges Gabriel Charpentier and David Samuel. 2023. https://doi.org/10.18653/v1/2023.conll-babylm.20 Not all layers are equally as important: Every layer counts BERT . In Proceedings of the BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, pages 238--252, Singapore. Association for Computational Linguistics

doi:10.18653/v1/2023.conll-babylm.20 2023
[5]

Hu, Tal Linzen, Aaron Mueller, Candance Ross, Alex Warstadt, Ethan Wilcox, Adina Williams, and Chengxu Zhuang

Leshem Choshen, Ryan Cotterell, Michael Y. Hu, Tal Linzen, Aaron Mueller, Candance Ross, Alex Warstadt, Ethan Wilcox, Adina Williams, and Chengxu Zhuang. 2024. https://babylm.github.io/2024-cfp.pdf Call for papers -- the babylm challenge: Sample-efficient pretraining on a developmentally plausible corpus

work page 2024
[6]

Judit Fazekas, Andrew Jessop, Julian Pine, and Caroline Rowland. 2020. https://doi.org/10.1098/rsos.180877 Do children learn from their prediction mistakes? a registered report evaluating error-based theories of language acquisition . Royal Society Open Science, 7(11):180877

work page doi:10.1098/rsos.180877 2020
[7]

Xudong Hong, Sharid Lo \'a iciga, and Asad Sayeed. 2023. https://doi.org/10.18653/v1/2023.conll-babylm.22 A surprisal oracle for active curriculum language modeling . In Proceedings of the BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, pages 259--268, Singapore. Association for Computational Linguistics

work page doi:10.18653/v1/2023.conll-babylm.22 2023
[8]

Ivanova, Aalok Sathe, Benjamin Lipkin, Unnathi Kumar, Setayesh Radkani, Thomas H

Anna A. Ivanova, Aalok Sathe, Benjamin Lipkin, Unnathi Kumar, Setayesh Radkani, Thomas H. Clark, Carina Kauf, Jennifer Hu, R. T. Pramod, Gabriel Grand, Vivian Paulun, Maria Ryskina, Ekin Akyurek, Ethan Wilcox, Nafisa Rashid, Leshem Choshen, Roger Levy, Evelina Fedorenko, Joshua Tenenbaum, and Jacob Andreas. 2024. https://arxiv.org/abs/2405.09605 Elements ...

arXiv 2024

Show all 17 references

[9]

Borna Jafarpour, Dawn Sepehr, and Nick Pogrebnyakov. 2021. https://doi.org/10.18653/v1/2021.internlp-1.6 Active curriculum learning . In Proceedings of the First Workshop on Interactive Learning for Natural Language Processing, pages 40--45, Online. Association for Computation...

2021 doi
[10]

Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. http://arxiv.org/abs/1907.11692 Roberta: A robustly optimized bert pretraining approach

2019 arXiv
[11]

Masek, Brianna T.M

Lillian R. Masek, Brianna T.M. McMillan, Sarah J. Paterson, Catherine S. Tamis-LeMonda, Roberta Michnick Golinkoff, and Kathy Hirsh-Pasek. 2021. https://doi.org/https://doi.org/10.1016/j.dr.2021.100961 Where language meets attention: How contingent interactions promote learnin...

2021
[12]

David Samuel, Andrey Kutuzov, Lilja vrelid, and Erik Velldal. 2023. https://aclanthology.org/2023.findings-eacl.146 Trained on 100 million words and still in shape: BERT meets B ritish N ational C orpus . In Findings of the Association for Computational Linguistics: EACL 2023,...

2023
[13]

Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R. Bowman. 2018. GLUE : A multi-task benchmark and analysis platform for natural language understanding. ArXiv preprint 1804.07461

2018 arXiv
[14]

Alex Warstadt, Aaron Mueller, Leshem Choshen, Ethan Gotlieb Wilcox, Chengxu Zhuang, Juan Ciro, Rafael Mosquera, Adina Williams, Bhargavi Paranjabe, Tal Linzen, and Ryan Cotterell. 2023. Findings of the 2023 B aby LM C hallenge: S ample-efficient pretraining on developmentally ...

2023
[15]

Alex Warstadt, Alicia Parrish, Haokun Liu, Anhad Mohananey, Wei Peng, Sheng-Fu Wang, and Samuel R. Bowman. 2020. https://doi.org/10.1162/tacl\_a\_00321 Blimp: The benchmark of linguistic minimal pairs for english . Transactions of the Association for Computational Linguistics,...

2020 doi
[16]

Zhisong Zhang, Emma Strubell, and Eduard Hovy. 2022. https://doi.org/10.18653/v1/2022.emnlp-main.414 A survey of active learning for natural language processing . In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 6166--6190, Abu D...

2022 doi
[17]

Liu Zhuang, Lin Wayne, Shi Ya, and Zhao Jun. 2021. https://aclanthology.org/2021.ccl-1.108 A robustly optimized BERT pre-training approach with post-training . In Proceedings of the 20th Chinese National Conference on Computational Linguistics, pages 1218--1227, Huhhot, China....

2021

Pith tools

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

[1] [1]

URL: " 'urlintro :=

ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

[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]

Yoshua Bengio, J\' e r\^ o me Louradour, Ronan Collobert, and Jason Weston. 2009. https://doi.org/10.1145/1553374.1553380 Curriculum learning . In Proceedings of the 26th Annual International Conference on Machine Learning, ICML '09, page 41–48, New York, NY, USA. Association for Computing Machinery

arXiv 2009

[4] [4]

Lucas Georges Gabriel Charpentier and David Samuel. 2023. https://doi.org/10.18653/v1/2023.conll-babylm.20 Not all layers are equally as important: Every layer counts BERT . In Proceedings of the BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, pages 238--252, Singapore. Association for Computational Linguistics

doi:10.18653/v1/2023.conll-babylm.20 2023

[5] [5]

Hu, Tal Linzen, Aaron Mueller, Candance Ross, Alex Warstadt, Ethan Wilcox, Adina Williams, and Chengxu Zhuang

Leshem Choshen, Ryan Cotterell, Michael Y. Hu, Tal Linzen, Aaron Mueller, Candance Ross, Alex Warstadt, Ethan Wilcox, Adina Williams, and Chengxu Zhuang. 2024. https://babylm.github.io/2024-cfp.pdf Call for papers -- the babylm challenge: Sample-efficient pretraining on a developmentally plausible corpus

work page 2024

[6] [6]

Judit Fazekas, Andrew Jessop, Julian Pine, and Caroline Rowland. 2020. https://doi.org/10.1098/rsos.180877 Do children learn from their prediction mistakes? a registered report evaluating error-based theories of language acquisition . Royal Society Open Science, 7(11):180877

work page doi:10.1098/rsos.180877 2020

[7] [7]

Xudong Hong, Sharid Lo \'a iciga, and Asad Sayeed. 2023. https://doi.org/10.18653/v1/2023.conll-babylm.22 A surprisal oracle for active curriculum language modeling . In Proceedings of the BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, pages 259--268, Singapore. Association for Computational Linguistics

work page doi:10.18653/v1/2023.conll-babylm.22 2023

[8] [8]

Ivanova, Aalok Sathe, Benjamin Lipkin, Unnathi Kumar, Setayesh Radkani, Thomas H

Anna A. Ivanova, Aalok Sathe, Benjamin Lipkin, Unnathi Kumar, Setayesh Radkani, Thomas H. Clark, Carina Kauf, Jennifer Hu, R. T. Pramod, Gabriel Grand, Vivian Paulun, Maria Ryskina, Ekin Akyurek, Ethan Wilcox, Nafisa Rashid, Leshem Choshen, Roger Levy, Evelina Fedorenko, Joshua Tenenbaum, and Jacob Andreas. 2024. https://arxiv.org/abs/2405.09605 Elements ...

arXiv 2024