Pith. sign in

REVIEW 3 major objections 6 minor 88 references

Towards Developmentally Plausible Rewards: Communicative Success as a Learning Signal for Interactive Language Models

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that communicative success in a language-only question-answering reference game can serve as a learning signal for language models, that the signal is at least indirectly about grammaticality, and that channel…

desk verdict An honest proof-of-concept for communicative-success rewards in a language-only reference game, with solid behavioral findings under bottlenecks but a grammaticality claim that is only partially supported. read the letter →

arxiv 2505.05970 v1 pith:DR3DM65N submitted 2025-05-09 cs.CL

Lennart Stöpler , Rufat Asadli , Mitja Nikolaus , Ryan Cotterell , Alex Warstadt This is my paper · ORCID
classification cs.CL
keywords communicativesuccessreinforcementlearninglanguageacquisitionreferencegamesummarizationgrammaticalitysurprisalquestionanswering
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 proposes that an artificial language model can learn from the simple fact of being understood, the way a child does. In the authors' 'abstract reference game,' a speaker summarizes a passage, a frozen question-answering listener answers a question from that summary, and the speaker is rewarded by how well the listener's answer matches a ground-truth answer. A feasibility study shows that corrupting the passage—truncating it, deleting stop words, or scrambling word order—degrades the listener's answers, so the reward carries at least an indirect signal about text quality and plausibly about grammaticality. Reinforcement-learning fine-tuning with this reward changes behavior: a length bottleneck drives summaries toward telegraphic, function-word-poor speech, while a surprisal bottleneck produces output closer to the original model's grammar. The paper reports no improvement in grammatical ability on the BLiMP and LanguageTool benchmarks, and training from scratch degenerates into nonsense.

What carries the argument

The mechanism is the abstract reference game: a speaker sees a passage, produces a summary, a frozen pretrained question-answering listener reads the summary and answers a question, and the reward is the ROUGE-L F1 score between the listener's answer and the ground truth. To keep the speaker from simply copying the passage, the score subtracts a penalty proportional either to the summary's length relative to the passage or to the listener's surprisal (negative log probability) of the summary relative to the passage. The speaker is trained with Proximal Policy Optimization, and the balance between reward and penalty is controlled by a single hyperparameter $\lambda$ through $\text{score} = (1-\lambda)\cdot\text{reward} - \lambda\cdot\text{penalty}$. This design converts the abstract goal of 'being understood' into a concrete, optimizable numerical signal, and lets the authors attribute changes in speaker behavior to specific communicative costs.

What would settle it

Prepare pairs of summaries that use exactly the same words but differ only in word order—one grammatical, one scrambled—and measure the ROUGE-L reward the listener returns; if the reward does not drop when grammar is destroyed, the signal carries no grammatical information. A complementary experiment is to train the speaker with a listener that ignores word order entirely and check whether BLiMP scores still fail to move.

Watch

Extended reading notes

Core claim

The central claim is that communicative success in a language-only question-answering reference game is a usable learning signal for language models, and specifically 'an indirect signal about grammaticality.' The authors make the case with a feasibility study: when the listener is fed ungrammatical or incomplete summaries, its answer quality degrades smoothly, so a speaker optimizing that reward has some incentive to produce grammatical, complete text. They then fine-tune a pretrained speaker with reinforcement learning, with a penalty that either rewards brevity or rewards low listener surprisal. The result is that the bottleneck changes what the speaker says in interpretable ways—copying under no penalty, telegraphic speech under a length penalty, and more conservative grammatical output under a surprisal penalty. None of these conditions improves grammatical knowledge on BLiMP or LanguageTool, and the paper presents that negative result as an open challenge rather than a refutation of the method.

Load-bearing premise

The load-bearing premise is that the frozen question-answering listener's answer quality, scored by ROUGE-L against a ground-truth answer, is a faithful proxy for human communicative success and, in particular, for grammaticality rather than for general text quality.

Editorial extensions

If this is right

  • If the reward is a valid learning signal, language models can in principle improve their output with no explicit grammatical feedback, purely from observing whether a listener understood them.
  • A length bottleneck reliably pushes outputs toward telegraphic summaries that drop function words, which could be used to model the one- and two-word utterance phases of child speech.
  • A surprisal bottleneck keeps summaries closer to the grammatical style of the base model, indicating that information-density pressure is more compatible with grammatical output than raw brevity is.
  • The lack of any BLiMP or LanguageTool improvement means the current recipe is missing at least one ingredient—longer training, a larger question-answering data budget, a different model, or a richer notion of communicative success.
  • Training from scratch with this reward fails outright, so the method cannot yet explain how a learner with no prior language bootstraps from interaction alone.

Reading between the lines

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

  • Because the feasibility study's corruptions change syntax and semantics together, the same evidence is consistent with the reward tracking general text quality rather than grammaticality specifically—a distinction the paper does not resolve.
  • A direct test this paper does not run would replace the frozen listener with a bag-of-words or order-insensitive reader; if grammaticality still influences the reward, word order is not the carrier, and if it stops, the signal is about syntax.
  • The 240 training steps and roughly 500,000-token fine-tuning budget are smaller than the scale at which interactive rewards might be expected to show grammatical gains, so the negative BLiMP result is more a ceiling on this configuration than on the method.
  • A curriculum that starts with a tight length bottleneck and gradually relaxes it could turn the observed telegraphic drift into a developmental trajectory, mirroring children's progression from one-word utterances to full sentences.
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.

Referee Report

3 major / 6 minor

Summary. This paper proposes the 'abstract reference game,' a language-only variant of Lewis-style reference games in which a speaker T5 model summarizes a passage, a frozen UnifiedQA listener answers a question based on the summary, and the speaker receives a reward based on ROUGE-L between the listener's answer and a ground-truth answer. The authors present a feasibility study (Section 5) showing that listener QA performance degrades under truncation, deletion, and scrambling of the context, which they interpret as evidence that the reward is an indirect signal for grammaticality. They then apply PPO to fine-tune T5 under length- or surprisal-based bottlenecks (Section 7) and find that a length bottleneck yields telegraphic, function-word-depleted output while a surprisal bottleneck preserves grammatical structure more; no improvement on BLiMP or LanguageTool is observed. Training from scratch (Section 6) fails to increase reward. The paper is honest about its negative results and frames the contribution as a proof of concept.

Significance. The abstract reference game is a novel and reasonably clean operationalization of communicative feedback for language acquisition research, and the use of a frozen listener to avoid semantic drift is well motivated. The paper reports controlled experiments with three seeds and multiple metrics, and it explicitly acknowledges the limitations of its training budget and benchmarks. If the feasibility claim were established, the paradigm would be a useful testbed for studying interaction as a learning signal in developmentally plausible LMs. However, the current evidence does not isolate grammaticality from content preservation, so the central claim is not yet supported. The observed bottleneck-dependent behavioral changes are interesting and reproducible, but their connection to grammatical learning remains speculative.

major comments (3)
  1. [Section 5, Figure 2] The feasibility study confounds grammaticality with content preservation. Truncation, deletion, and word-order scrambling all remove or rearrange semantic information, so the observed degradation in listener performance is expected even if the listener is indifferent to grammar. The only syntax-specific manipulation, stop-word removal, is reported as a 'minor loss' with no error bars or significance test. To support the abstract's claim that the reward is an 'indirect signal about grammaticality,' the authors should test matched-content minimal pairs (e.g., grammatical vs. ungrammatical paraphrases with identical propositions) and show that the reward separates them. As it stands, the gradient in Figure 2 may reflect general text quality rather than grammar specifically.
  2. [Section 5, Figure 2] The figure reports macro-averages over 17 datasets without error bars, confidence intervals, or statistical tests. Claims such as 'minor loss' and 'inversely correlated' rest on visual inspection of single points. Because the feasibility study is the primary evidence for the central claim, the authors should provide variation across datasets and/or repeated perturbation draws, or at least a table of per-condition values with standard deviations.
  3. [Section 4, Equations for SCORElength and SCOREsurprisal] The reward is ROUGE-L between the listener's answer and the ground truth. ROUGE-L is a lexical overlap metric that is not validated as a proxy for human communicative success or for grammaticality. The paper should at least discuss this limitation, and ideally provide a small human evaluation or a correlation with an alternative metric (e.g., accuracy on multiple-choice QA subsets) to justify the operationalization.
minor comments (6)
  1. [Section 2.1] The phrase 'GPTs, Llamas, and Pythias' should read 'GPTs, Llamas, and Pythia' (or 'Pythia models').
  2. [Section 3] The citation 'V ong et al.' contains a spacing artifact and should be 'Vong et al.'.
  3. [References] Several references contain spacing artifacts, e.g., 'F orm and Function' for Bloom (1970). These should be cleaned up.
  4. [Figure 2] The caption says 'T5 QA Performance', but the listener is UnifiedQA; rename to 'QA listener performance' to avoid confusion.
  5. [Section 5] The specific values of the perturbation parameters (c, m, s, d) for the low/medium/high conditions are not provided, which makes the feasibility study not fully reproducible; please include them in the appendix.
  6. [Section 7] The fine-tuning experiments use only 240 PPO steps on a 540k-token SQuAD 2.0 validation split. The absence of grammatical improvements is therefore unsurprising and should not be overinterpreted; the authors' discussion already acknowledges this, but the point could be stated even more explicitly in the abstract.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reward is defined externally via QA/ROUGE, and grammaticality enters only as an empirical probe.

full rationale

The paper's reward is exogenously defined as ROUGE-L between a frozen UnifiedQA listener's answer and a ground-truth answer (Section 4 score equations), with grammaticality nowhere in the objective. The 'grammaticality signal' claim is tested empirically in Section 5 by perturbing contexts and measuring listener degradation; this is a falsifiable probe rather than a fitted input or a self-referential definition. The training experiments are evaluated against external benchmarks (BLiMP, LanguageTool), and the paper reports the null result that grammar does not improve, which is itself inconsistent with a circular design. The bottleneck penalties (length and listener surprisal) are separate, interpretable regularizers rather than renamed predictions. Self-citations (Warstadt et al.; Nikolaus & Fourtassi) provide background and evaluation tools, but none is load-bearing: BLiMP is independently defined, and the feedback typology is not used to derive the results. The feasibility study's confound between syntax and semantics is a validity limitation, not circularity, because the reward is not defined in terms of grammaticality; the paper never reduces its central claim to an identity with its inputs.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

No new physical or mathematical entities are introduced. The abstract reference game is a task rather than an entity. All quantities derive from standard language modeling, reinforcement learning, and question-answering components.

free parameters (2)
  • lambda (bottleneck weight) = 0, 0.1, 0.5, 0.9, 1.0
    Hand-chosen across five values; every reported trend is conditional on this sweep, so it is a selected design parameter even though no single value is fitted to a target outcome.
  • PPO training hyperparameters = learning rate 1.41e-5, PPO epochs 4, batch size 512, top-k 50, top-p 0.1, beams 3
    Selected by a Bayesian sweep in the appendix; speaker behavior and training stability depend on this configuration.
assumptions (5)
  • domain assumption QA accuracy of the frozen UnifiedQA listener is a valid measure of communicative success.
    The training reward is ROUGE between the listener's answer and the ground truth (Section 4), so the entire method assumes this proxy is meaningful.
  • domain assumption ROUGE-L F1 is an adequate semantic similarity metric for short answers.
    ROUGE is lexical overlap, so it can miss semantically correct paraphrases (Section 4).
  • domain assumption Listener surprisal is a valid proxy for comprehension difficulty.
    Used to define the surprisal bottleneck in Section 3.2; the paper follows Shannon and Levy but relies on a neural LM estimate.
  • standard math PPO is stable enough for training the speaker in this small-scale setting.
    The training from scratch experiment degenerates (Section 6), so this assumption fails in at least that setting; fine-tuning is more stable.
  • ad hoc to paper Text perturbations in the feasibility study isolate grammaticality from other forms of degradation.
    Stop-word removal, truncation, and scrambling change semantics and length as well as grammar, so the grammaticality-signal conclusion is indirect (Section 5).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Developmentally Plausible Rewards: Communicative Success as a Learning Signal for Interactive Language Models." pith.science (2026). https://pith.science/paper/DR3DM65N

@misc{pith2026250505970,
  author       = {Pith},
  title        = {Pith review of: Towards Developmentally Plausible Rewards: Communicative Success as a Learning Signal for Interactive Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DR3DM65N}},
  note         = {Machine review of arXiv:2505.05970}
}
read the original abstract

We propose a method for training language models in an interactive setting inspired by child language acquisition. In our setting, a speaker attempts to communicate some information to a listener in a single-turn dialogue and receives a reward if communicative success is achieved. Unlike earlier related work using image--caption data for interactive reference games, we operationalize communicative success in a more abstract language-only question--answering setting. First, we present a feasibility study demonstrating that our reward provides an indirect signal about grammaticality. Second, we conduct experiments using reinforcement learning to fine-tune language models. We observe that cognitively plausible constraints on the communication channel lead to interpretable changes in speaker behavior. However, we do not yet see improvements on linguistic evaluations from our training regime. We outline potential modifications to the task design and training configuration that could better position future work to use our methodology to observe the benefits of interaction on language learning in computational cognitive models.

Figures

Figures reproduced from arXiv: 2505.05970 by the authors.

Figure 1
Figure 1. Illustration of the summarization game. two reasons: First, LMs have the potential to transform computational modeling of human language processing and acquisition (Warstadt & Bowman, 2022), but they are currently too divergent from humans to meet their full potential (Baroni, 2022). Second, humans are orders of magnitude more data-efficient at acquiring language than LMs (Warstadt et al., 2023), and this may be due… view at source ↗
Figure 3
Figure 3. The trajectory of reward, penalty, and score, as well as grammatical performance during [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 2
Figure 2. Results from the feasibility study. y￾axis shows macro average performance over the 17 datasets from UnifiedQA (Khashabi et al., 2022). Our objective is mainly to provide a proof of con￾cept that an abstract reference game provides a learning signal for language acquisition in LMs. To validate the viability of our setup in princi￾ple, we conduct a feasibility study measuring how the response of the listener model ch… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Example reward trajectories in models trained from scratch, with rolling average window of [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

88 extracted references · 45 canonical work pages

  1. [1]

    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 format.date year duplicate empty "emp...

  2. [2]

    Lmrl gym: Benchmarks for multi-turn reinforcement learning with language models, 2023

    Marwa Abdulhai, Isadora White, Charlie Snell, Charles Sun, Joey Hong, Yuexiang Zhai, Kelvin Xu, and Sergey Levine. Lmrl gym: Benchmarks for multi-turn reinforcement learning with language models, 2023

    2023

  3. [3]

    On the proper role of linguistically-oriented deep net analysis in linguistic theorizing

    Marco Baroni. On the proper role of linguistically-oriented deep net analysis in linguistic theorizing. In Shalom Lappin and Jean Philippe Bernardy (eds.), Algebraic Structures in Natural Language, pp.\ 1--16. CRC Press, 2022

    2022

  4. [4]

    Pythia: a suite for analyzing large language models across training and scaling

    Stella Biderman, Hailey Schoelkopf, Quentin Gregory Anthony, Herbie Bradley, Kyle O'Brien, Eric Hallahan, Mohammad Aflah Khan, Shivanshu Purohit, Usvsn Sai Prashanth, Edward Raff, Aviya Skowron, Lintang Sutawika, and Oskar Van Der Wal. Pythia: a suite for analyzing large language models across training and scaling. In Andreas Krause, Emma Brunskill, Kyung...

    2023

  5. [5]

    Experiment Tracking with Weights and Biases , 2020

    Lukas Biewald. Experiment Tracking with Weights and Biases , 2020. URL https://www.wandb.com/

    2020

  6. [6]

    Language Development : Form and Function in Emerging Grammars

    Lois Bloom. Language Development : Form and Function in Emerging Grammars . MIT Press, 1970

    1970

  7. [7]

    Derivational complexity and order of acquisition in child speech

    Roger Brown and Camille Hanlon. Derivational complexity and order of acquisition in child speech. Cognition and the development of language, 1970

    1970

  8. [8]

    Child's Talk : Learning to Use Language

    Jerome Bruner. Child's Talk : Learning to Use Language . Child Language Teaching and Therapy, 1 0 (1): 0 111--114, 1985. ISSN 0265-6590, 1477-0865. doi:10.1177/026565908500100113

    work page doi:10.1177/026565908500100113 1985

Show all 88 references
  1. [9]

    Meaning and necessity

    Rudolf Carnap. Meaning and necessity. University of Chicago Press, 1947

    1947

  2. [10]

    Chang and Benjamin K

    Tyler A. Chang and Benjamin K. Bergen. Word Acquisition in Neural Language Models . Transactions of the Association for Computational Linguistics, 10: 0 1--16, January 2022. ISSN 2307-387X. doi:10.1162/tacl_a_00444. URL https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00...

    2022 doi

  3. [11]

    Conceptual 12M : Pushing web-scale image-text pre-training to recognize long-tail visual concepts

    Soravit Changpinyo, Piyush Sharma, Nan Ding, and Radu Soricut. Conceptual 12M : Pushing web-scale image-text pre-training to recognize long-tail visual concepts. In 2021 IEEE / CVF conference on computer vision and pattern recognition ( CVPR ) , pp.\ 3557--3567, June 2021. doi...

    2021

  4. [12]

    Not all layers are equally as important: Every Layer Counts BERT

    Lucas Georges Gabriel Charpentier and David Samuel. Not all layers are equally as important: Every Layer Counts BERT . In Proceedings of the BabyLM Challenge . Association for Computational Linguistics (ACL), December 2023

    2023

  5. [13]

    Chouinard and Eve V

    Michelle M. Chouinard and Eve V. Clark. Adult reformulations of child errors as negative evidence. Journal of Child Language, 30 0 (3): 0 637--669, 2003. ISSN 1469-7602, 0305-0009. doi:10.1017/S0305000903005701

    2003 doi

  6. [14]

    Eve V. Clark. Conversation and Language Acquisition : A Pragmatic Approach . Language Learning and Development, 14 0 (3): 0 170--185, 2018. ISSN 1547-5441. doi:10.1080/15475441.2017.1340843

    2018

  7. [15]

    Eve V. Clark. Conversational Repair and the Acquisition of Language . Discourse Processes, 57 0 (5-6): 0 441--459, 2020. ISSN 0163-853X. doi:10.1080/0163853X.2020.1719795

    2020

  8. [16]

    Automatic text summarization: A comprehensive survey

    Wafaa S El-Kassas, Cherif R Salama, Ahmed A Rafea, and Hoda K Mohamed. Automatic text summarization: A comprehensive survey. Expert systems with applications, 165: 0 113679, 2021

    2021

  9. [17]

    Johnson, Annette Karmiloff-Smith, Domenico Parisi, and Kim Plunkett

    Jeffrey Elman, Elizabeth Bates, Mark H. Johnson, Annette Karmiloff-Smith, Domenico Parisi, and Kim Plunkett. Rethinking innateness: a connectionist perspective on development. The MIT Press, October 1996. ISBN 978-0-262-27229-2. doi:10.7551/mitpress/5929.001.0001. URL https://...

    1996 doi

  10. [18]

    Finding structure in time

    Jeffrey L Elman. Finding structure in time. Cognitive science, 14 0 (2): 0 179--211, 1990. Publisher: Wiley Online Library

    1990

  11. [19]

    A distributional perspective on word learning in neural language models

    Filippo Ficarra, Ryan Cotterell, and Alex Warstadt. A distributional perspective on word learning in neural language models. arXiv preprint arXiv:2502.05892, 2025

    2025 arXiv

  12. [20]

    Michael C. Frank. Bridging the data gap between children and large language models. Trends in Cognitive Sciences, 27 0 (11): 0 990--992, November 2023. ISSN 1364-6613, 1879-307X. doi:10.1016/j.tics.2023.08.007. URL https://www.cell.com/trends/cognitive-sciences/abstract/S1364-...

    2023 doi

  13. [21]

    Richards, Steven F

    Jill Gilkerson, Jeffrey A. Richards, Steven F. Warren, Judith K. Montgomery, Charles R. Greenwood, D. Kimbrough Oller, John H. L. Hansen, and Terrance D. Paul. Mapping the Early Language Environment Using All - Day Recordings and Automated Analysis . American Journal of Speech...

    2017 doi

  14. [22]

    Towards pragmatic production strategies for natural language generation tasks, 2022

    Mario Giulianelli. Towards pragmatic production strategies for natural language generation tasks, 2022

    2022

  15. [23]

    Language identification in the limit

    E Mark Gold. Language identification in the limit. Information and Control, 10 0 (5): 0 447--474, May 1967. ISSN 0019-9958. doi:10.1016/S0019-9958(67)91165-5. URL https://www.sciencedirect.com/science/article/pii/S0019995867911655

    1967 doi

  16. [24]

    H. P. Grice. Logic and Conversation . Brill, December 1975. ISBN 978-90-04-36881-1. doi:10.1163/9789004368811_003. Pages: 41-58 Section: Speech Acts

    1975 doi

  17. [25]

    Dynamic population-based meta-learning for multi-agent communication with natural language

    Abhinav Gupta, Marc Lanctot, and Angeliki Lazaridou. Dynamic population-based meta-learning for multi-agent communication with natural language. In Advances in Neural Information Processing Systems, volume 34, pp.\ 16899--16912, 2021

    2021

  18. [26]

    A Data -driven Investigation of Corrective Feedback on Subject Omission Errors in First Language Acquisition

    Sarah Hiller and Raquel Fernandez. A Data -driven Investigation of Corrective Feedback on Subject Omission Errors in First Language Acquisition . In Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning , pp.\ 105--114, Berlin, Germany, 2016. Ass...

    2016 doi

  19. [27]

    Yanping Huang and Rajesh P. N. Rao. Predictive coding. WIREs Cognitive Science, 2 0 (5): 0 580--593, September 2011. ISSN 1939-5078, 1939-5086. doi:10.1002/wcs.142. URL https://wires.onlinelibrary.wiley.com/doi/10.1002/wcs.142

    2011 doi

  20. [28]

    Multitasking inhibits semantic drift, 2021

    Athul Paul Jacob, Mike Lewis, and Jacob Andreas. Multitasking inhibits semantic drift, 2021

    2021

  21. [29]

    Why Large Language Models Are Poor Theories of Human Linguistic Cognition : A Reply to Piantadosi

    Roni Katzir. Why Large Language Models Are Poor Theories of Human Linguistic Cognition : A Reply to Piantadosi . Biolinguistics, 17 0 (Article e13153), 2023. doi:https://doi.org/10.5964/bioling.13153. URL https://bioling.psychopen.eu/index.php/bioling/article/view/13153

    2023 doi

  22. [30]

    Unifiedqa: Crossing format boundaries with a single qa system, 2020

    Daniel Khashabi, Sewon Min, Tushar Khot, Ashish Sabharwal, Oyvind Tafjord, Peter Clark, and Hannaneh Hajishirzi. Unifiedqa: Crossing format boundaries with a single qa system, 2020

    2020

  23. [31]

    Unifiedqa-v2: Stronger generalization via broader cross-format training, 2022

    Daniel Khashabi, Yeganeh Kordi, and Hannaneh Hajishirzi. Unifiedqa-v2: Stronger generalization via broader cross-format training, 2022

    2022

  24. [32]

    The handbook of pidgin and creole studies

    Silvia Kouwenberg and John Victor Singler. The handbook of pidgin and creole studies. John Wiley & Sons, 2009

    2009

  25. [33]

    Large language models and the argument from the poverty of the stimulus

    Nur Lan, Emmanuel Chemla, and Roni Katzir. Large language models and the argument from the poverty of the stimulus. Linguistic Inquiry, pp.\ 1--28, August 2024. ISSN 0024-3892. doi:10.1162/ling_a_00533. URL https://doi.org/10.1162/ling_a_00533

    2024 doi

  26. [34]

    Grammaticality, acceptability, and probability: A probabilistic view of linguistic knowledge

    Jey Han Lau, Alexander Clark, and Shalom Lappin. Grammaticality, acceptability, and probability: A probabilistic view of linguistic knowledge. Cognitive Science, 41 0 (5): 0 1202--1241, 2017. Publisher: Wiley Online Library

    2017

  27. [35]

    Multi-agent communication meets natural language: Synergies between functional and structural language learning, 2020 a

    Angeliki Lazaridou, Anna Potapenko, and Olivier Tieleman. Multi-agent communication meets natural language: Synergies between functional and structural language learning, 2020 a

    2020

  28. [36]

    Multi-agent Communication meets Natural Language : Synergies between Functional and Structural Language Learning

    Angeliki Lazaridou, Anna Potapenko, and Olivier Tieleman. Multi-agent Communication meets Natural Language : Synergies between Functional and Structural Language Learning . In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , pp.\ 7663--...

    2020 doi

  29. [37]

    Expectation-based syntactic comprehension

    Roger Levy. Expectation-based syntactic comprehension. Cognition, 106 0 (3): 0 1126--1177, March 2008. ISSN 00100277. doi:10.1016/j.cognition.2007.05.006. URL https://linkinghub.elsevier.com/retrieve/pii/S0010027707001436

    2008 doi

  30. [38]

    Florian Jaeger

    Roger Levy and T. Florian Jaeger. Speakers optimize information density through syntactic reduction. In Bernhard Schölkopf, John Platt, and Thomas Hofmann (eds.), Advances in Neural Information Processing Systems 19 , pp.\ 849--856. The MIT Press, September 2007. ISBN 978-0-26...

    2007 doi

  31. [39]

    David K. Lewis. On the Plurality of Worlds. Wiley-Blackwell, Malden, Mass., 1986

    1986

  32. [40]

    Convention: A Philosophical Study

    David Kellogg Lewis. Convention: A Philosophical Study. Wiley-Blackwell, Cambridge, MA, USA, 1969

    1969

  33. [41]

    ROUGE : A package for automatic evaluation of summaries

    Chin-Yew Lin. ROUGE : A package for automatic evaluation of summaries. In Text Summarization Branches Out, pp.\ 74--81, Barcelona, Spain, July 2004. Association for Computational Linguistics. URL https://aclanthology.org/W04-1013

    2004

  34. [42]

    What can linguistics and deep learning contribute to each other? Response to Pater

    Tal Linzen. What can linguistics and deep learning contribute to each other? Response to Pater . Language, 95 0 (1): 0 e99--e108, 2019. Publisher: Linguistic Society of America

    2019

  35. [43]

    Computational Language Acquisition with Theory of Mind , March 2023

    Andy Liu, Hao Zhu, Emmy Liu, Yonatan Bisk, and Graham Neubig. Computational Language Acquisition with Theory of Mind , March 2023. URL http://arxiv.org/abs/2303.01502. arXiv:2303.01502 [cs]

    2023 arXiv

  36. [44]

    On the interaction between supervision and self-play in emergent communication, 2020

    Ryan Lowe, Abhinav Gupta, Jakob Foerster, Douwe Kiela, and Joelle Pineau. On the interaction between supervision and self-play in emergent communication, 2020

    2020

  37. [45]

    Countering language drift with seeded iterated learning, 2020

    Yuchen Lu, Soumye Singhal, Florian Strub, Olivier Pietquin, and Aaron Courville. Countering language drift with seeded iterated learning, 2020

    2020

  38. [46]

    Words and the World : Predictive Coding and the Language - Perception - Cognition Interface

    Gary Lupyan and Andy Clark. Words and the World : Predictive Coding and the Language - Perception - Cognition Interface . Current Directions in Psychological Science, 24 0 (4): 0 279--284, August 2015. ISSN 0963-7214, 1467-8721. doi:10.1177/0963721415570732. URL https://journa...

    2015 doi

  39. [47]

    Gary F. Marcus. Negative evidence in language acquisition. Cognition, 46 0 (1): 0 53--85, 1993. ISSN 0010-0277. doi:10.1016/0010-0277(93)90022-N

    1993 doi

  40. [48]

    CLIMB – curriculum learning for infant-inspired model building

    Richard Diehl Martinez, Hope McGovern, Zebulon Goriely, Christopher Davis, Andrew Caines, Paula Buttery, and Lisa Beinborn. CLIMB – curriculum learning for infant-inspired model building. In Proceedings of the BabyLM Challenge . Association for Computational Linguistics (ACL),...

    2023

  41. [49]

    The llama 3 herd of models

    AI @ Meta. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024. URL https://arxiv.org/abs/2407.21783

    2024 arXiv

  42. [50]

    Language Models as Models of Language

    Raphaël Millière. Language Models as Models of Language . In R. Nefdt, G. Dupre, and K. Stanton (eds.), The Oxford Handbook of the Philosophy of Linguistics . Oxford University Press, forthcoming. URL http://arxiv.org/abs/2408.07144. arXiv:2408.07144 [cs]

    arXiv

  43. [51]

    nanot5: A pytorch framework for pre-training and fine-tuning t5-style models with limited resources

    Piotr Nawrot. nanot5: A pytorch framework for pre-training and fine-tuning t5-style models with limited resources. ArXiv, abs/2309.02373, 2023

    2023 arXiv

  44. [52]

    Modeling the interaction between perception-based and production-based learning in children ' s early acquisition of semantic knowledge

    Mitja Nikolaus and Abdellah Fourtassi. Modeling the interaction between perception-based and production-based learning in children ' s early acquisition of semantic knowledge. In Arianna Bisazza and Omri Abend (eds.), Proceedings of the 25th Conference on Computational Natural...

    2021 doi

  45. [53]

    Communicative Feedback in Language Acquisition

    Mitja Nikolaus and Abdellah Fourtassi. Communicative Feedback in Language Acquisition . New Ideas in Psychology, 2023. doi:10.1016/j.newideapsych.2022.100985

    2023

  46. [54]

    Communicative Feedback as a Mechanism Supporting the Production of Intelligible Speech in Early Childhood

    Mitja Nikolaus, Laurent Prévot, and Abdellah Fourtassi. Communicative Feedback as a Mechanism Supporting the Production of Intelligible Speech in Early Childhood . In Proceedings of the 44th Annual Meeting of the Cognitive Science Society , 2022. doi:DOI: 10.31234/osf.io/sg5mv

    2022 doi

  47. [55]

    Communicative Feedback in Response to Children 's Grammatical Errors

    Mitja Nikolaus, Laurent Prévot, and Abdellah Fourtassi. Communicative Feedback in Response to Children 's Grammatical Errors . In Proceedings for the 45th Annual Meeting of the Cognitive Science Society , 2023

    2023

  48. [56]

    Language acquisition through language use: The functional sources of children's early utterances

    Anat Ninio and Catherine Snow. Language acquisition through language use: The functional sources of children's early utterances. Categories and Processes in Language Acquisition, 1988

    1988

  49. [57]

    ChatGPT : Optimizing language models for dialogue, 2022

    OpenAI. ChatGPT : Optimizing language models for dialogue, 2022. URL https://openai.com/blog/chatgpt/

    2022

  50. [58]

    Training language models to follow instructions with human feedback

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul F Christiano, Jan Leike, a...

    2022

  51. [59]

    Christiano, Jan Leike, and Ryan Lowe

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul F. Christiano, Jan Leike, ...

    2022

  52. [60]

    Bleu: a method for automatic evaluation of machine translation

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting on Association for Computational Linguistics, ACL '02, pp.\ 311–318, USA, 2002. Association for Computational ...

    2002

  53. [61]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer, 2023

    2023

  54. [62]

    Know what you don ' t know: Unanswerable questions for SQ u AD

    Pranav Rajpurkar, Robin Jia, and Percy Liang. Know what you don ' t know: Unanswerable questions for SQ u AD . In Iryna Gurevych and Yusuke Miyao (eds.), Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pp.\ 784-...

    2018 doi

  55. [63]

    Language evolution with deep learning

    Mathieu Rita, Paul Michel, Rahma Chaabouni, Olivier Pietquin, Emmanuel Dupoux, and Florian Strub. Language evolution with deep learning. arXiv preprint arXiv:2403.11958, 2024

    2024 arXiv

  56. [64]

    Information structure: Towards an integrated formal theory of pragmatics

    Craige Roberts. Information structure: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics, 5: 0 6--1, 1996

    1996

  57. [65]

    The Contrast Theory of negative input

    Matthew Saxton. The Contrast Theory of negative input. Journal of Child Language, 24 0 (1): 0 139--161, 1997. ISSN 1469-7602, 0305-0009. doi:10.1017/S030500099600298X

    1997 doi

  58. [66]

    Schegloff

    Emanuel A. Schegloff. Repair after next turn: The last structurally provided defense of intersubjectivity in conversation. American Journal of Sociology, 97 0 (5): 0 1295--1345, 1992. doi:10.1086/229903. URL https://doi.org/10.1086/229903. tex.eprint: https://doi.org/10.1086/229903

    1992 doi

  59. [67]

    Three Myths from the Language Acquisition Literature

    Ted Schoneberger. Three Myths from the Language Acquisition Literature . The Analysis of Verbal Behavior, 26 0 (1): 0 107--131, April 2010. ISSN 0889-9401, 2196-8926. doi:10.1007/BF03393086. URL http://link.springer.com/10.1007/BF03393086

    2010 doi

  60. [68]

    Proximal policy optimization algorithms, 2017

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms, 2017

    2017

  61. [69]

    Speech Acts : An Essay in the Philosophy of Language

    John Rogers Searle. Speech Acts : An Essay in the Philosophy of Language . Cambridge University Press, January 1969. ISBN 978-0-521-09626-3. Google-Books-ID: t3\_WhfknvF0C

    1969

  62. [70]

    Liu, and Christopher D

    Abigail See, Peter J. Liu, and Christopher D. Manning. Get to the point: Summarization with pointer-generator networks. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp.\ 1073--1083, Vancouver, Canada, July ...

    2017 doi

  63. [71]

    Children creating language: How nicaraguan sign language acquired a spatial grammar

    Ann Senghas and Marie Coppola. Children creating language: How nicaraguan sign language acquired a spatial grammar. Psychological Science, 12 0 (4): 0 323--328, 2001. doi:10.1111/1467-9280.00359. URL https://doi.org/10.1111/1467-9280.00359. tex.eprint: https://doi.org/10.1111/...

    2001

  64. [72]

    Dignum Sepulveda

    Krijn F. Dignum Sepulveda. Towards pragmatic efficiency in natural language generation: Optimizing communication costs and utility in summarization tasks. [Unpublished Bachelor's thesis], 2023

    2023

  65. [73]

    fMRI reveals language-specific predictive coding during naturalistic sentence comprehension

    Cory Shain, Idan Asher Blank, Marten van Schijndel, William Schuler, and Evelina Fedorenko. fMRI reveals language-specific predictive coding during naturalistic sentence comprehension. Neuropsychologia, 138: 0 107307, 2020. ISSN 0028-3932. doi:https://doi.org/10.1016/j.neurops...

    2020

  66. [74]

    C. E. Shannon. A Mathematical Theory of Communication . Bell System Technical Journal, 27 0 (3): 0 379--423, 1948. ISSN 1538-7305. doi:10.1002/j.1538-7305.1948.tb01338.x

    1948

  67. [75]

    Predictive coding: a fresh view of inhibition in the retina

    Mandyam Veerambudi Srinivasan, Simon Barry Laughlin, and Andreas Dubs. Predictive coding: a fresh view of inhibition in the retina. Proceedings of the Royal Society of London. Series B. Biological Sciences, 216 0 (1205): 0 427--459, 1982. Publisher: The Royal Society London

    1982

  68. [76]

    Stalnaker

    Robert C. Stalnaker. Assertion. Pragmatics, pp.\ 315--332, December 1978. doi:10.1163/9789004368873_013

    1978 doi

  69. [77]

    Constructing a Language : A Usage - Based Theory of Language Acquisition

    Michael Tomasello. Constructing a Language : A Usage - Based Theory of Language Acquisition . Harvard University Press, 2003. ISBN 978-0-674-01764-1

    2003

  70. [78]

    Emin Orhan, and Brenden M

    Wai Keen Vong, Wentao Wang, A. Emin Orhan, and Brenden M. Lake. Grounded language acquisition through the eyes and ears of a single child. Science, 383 0 (6682): 0 504--511, 2024. doi:10.1126/science.adi1374. URL https://www.science.org/doi/abs/10.1126/science.adi1374

    2024 doi

  71. [79]

    What artificial neural networks can tell us about human language acquisition

    Alex Warstadt and Samuel R Bowman. What artificial neural networks can tell us about human language acquisition. In Shalom Lappin and Jean-Philippe Bernardy (eds.), Algebraic Structures in Natural Language , pp.\ 17--60. CRC Press, 2022. Publisher: CRC Press

    2022

  72. [80]

    Alex Warstadt, Alicia Parrish, Haokun Liu, Anhad Mohananey, Wei Peng, Sheng-Fu Wang, and Samuel R. Bowman. BLiMP : The Benchmark of Linguistic Minimal Pairs for English . Transactions of the Association for Computational Linguistics, 8: 0 377--392, July 2020. ISSN 2307-387X

    2020

  73. [81]

    Call for papers -- the babylm challenge: Sample-efficient pretraining on a developmentally plausible corpus

    Alex Warstadt, Leshem Choshen, Aaron Mueller, Adina Williams, Ethan Wilcox, and Chengxu Zhuang. Call for papers -- the babylm challenge: Sample-efficient pretraining on a developmentally plausible corpus. Computing Research Repository, arXiv:2301.11796, 2023

    2023 arXiv

  74. [82]

    Using computational models to test syntactic learnability

    Ethan Gotlieb Wilcox, Richard Futrell, and Roger Levy. Using computational models to test syntactic learnability. Linguistic Inquiry, pp.\ 1--44, April 2023. ISSN 0024-3892. doi:10.1162/ling_a_00491. URL https://doi.org/10.1162/ling_a_00491

    2023 doi

  75. [83]

    A survey on knowledge distillation of large language models, 2024

    Xiaohan Xu, Ming Li, Chongyang Tao, Tao Shen, Reynold Cheng, Jinyang Li, Can Xu, Dacheng Tao, and Tianyi Zhou. A survey on knowledge distillation of large language models, 2024

    2024

  76. [84]

    A communicative approach to early word learning

    Daniel Yurovsky. A communicative approach to early word learning. New Ideas in Psychology, 50: 0 73--79, August 2018. ISSN 0732-118X. doi:10.1016/j.newideapsych.2017.09.001. URL https://www.sciencedirect.com/science/article/pii/S0732118X17300077

    2018 doi

  77. [85]

    Yian Zhang, Alex Warstadt, Xiaocheng Li, and Samuel R. Bowman. When Do You Need Billions of Words of Pretraining Data ? In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Pr...

    2021 doi

  78. [86]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  79. [87]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  80. [88]

    back channel

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

    2019

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.