Pith. sign in

REVIEW 3 major objections 6 minor 43 references

Mastering emergent language: learning to guide in simulated navigation

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read An autonomous Guide that communicates through a discrete emergent language speeds up learning for new navigation agents, generalizes across tasks, and is interpretable enough to be spoken back by the authors with 72-76% obedience.

desk verdict Solid speedup and transfer results in emergent navigation guidance, but the 'mastered the language' section needs causal controls before it can be believed. read the letter →

arxiv 1908.05135 v1 pith:X64ZG5MZ submitted 2019-08-14 cs.CL cs.AIcs.MA

Mathijs Mul , Diane Bouchacourt , Elia Bruni This is my paper · ORCID
classification cs.CLcs.AIcs.MA
keywords emergentcommunicationdiscreteimitationlearningBabyAInavigationinterpretabilityCausalInfluenceofgridworld
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 an autonomous Guide agent can learn, with no external supervision, a discrete communication protocol that helps a fresh Learner agent master navigation and manipulation tasks faster. On all six BabyAI levels tested, pairing a pretrained Guide with a new Learner sharply increases learning speed, and the Guide's messages transfer to harder levels the Guide never trained on. The paper further argues that the emergent language is highly interpretable: messages correlate with the Learner's actions and with salient events in the environment, and the authors can replay choreographed messages to reach 72-76% obedience. If this holds, emergent and unsupervised communication can act as a reusable, interactive alternative to scripted instruction for sequential decision-making agents.

What carries the argument

The load-bearing mechanism is the Guide's discrete communication bottleneck: between the Guide's memory LSTM and its policy, a decoder must compress continuous information into a sequence of discrete tokens, made trainable with a straight-through Gumbel softmax estimator. Because the message length is fixed at two and the vocabulary at three, every piece of guidance must pass through one of nine messages, which forces the protocol to be compact and reusable. The Causal Influence of Communication metric then measures how much the Learner's next action depends on the message, and correlation analysis over message-action and message-observation pairs supplies the interpretation the authors use to send their own commands.

What would settle it

In GoToObj and GoToLocal, compare the Learner's obedience to the same preset messages under three conditions: normal visual input, masked visual input, and random messages. If the 72-76% obedience persists when visual input is masked, the messages themselves carry the instruction; if obedience collapses, the reported numbers partly reflect the environment and the Learner's own policy.

Watch

Extended reading notes

Core claim

Working in the BabyAI gridworld, the paper trains a Guide whose memory output must pass through a discrete bottleneck and be emitted as a fixed-length message of two tokens drawn from a vocabulary of three, giving nine possible messages. A pretrained Guide is then coupled with a new Learner, and both are trained by behavioral cloning on demonstrations from reinforcement-learned experts. The central result is that the Guide's messages cause a large increase in the Learner's learning speed on every level tested, that a Guide pretrained on one level still helps at related unseen levels, and that a Guide trained on three levels at once is as helpful as a level-specialized Guide. The Causal Influence of Communication metric shows the messages matter most early in training and fade as the Learner becomes independent. Using message-action correlations computed at the highest-influence epoch, the authors hijack the Guide and send preset messages, successfully directing a pirouette and a waltz and obtaining 72-76% obedience.

Load-bearing premise

The speech-back claim assumes that message-action correlations measured on training-distribution frames stay valid when the authors hijack the Guide and send preset messages out of distribution, while the Learner still receives its own visual observations.

Editorial extensions

If this is right

  • On all six BabyAI levels tested, a new Learner paired with a pretrained Guide reaches high validation success faster than an unguided baseline Learner.
  • Guidance learned on one level transfers to related harder levels, so the messages encode generalizable skills instead of level-specific solutions.
  • A single Guide trained on three levels at once is as helpful as per-level specialized Guides, showing the protocol is not overfit to one task.
  • The causal influence of messages is highest early in training and declines as the Learner improves, so guidance acts as a temporary scaffold rather than a permanent dependency.
  • Message-action correlations are strong enough that preset choreographed messages produce obedience rates of 72-76%, letting an outside party address the Learner in its own language.

Reading between the lines

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

  • The 72-76% obedience figures should be read as an upper bound on linguistic control: because the Learner also sees the grid, an ablation that masks visual input or scrambles messages would separate message-driven compliance from environment-driven behavior.
  • Because the Guide is finetuned while paired with the Learner, the protocol is co-negotiated rather than fixed; freezing the pretrained Guide would test whether the language remains stable and reusable across different Learners.
  • The CIC decline suggests a testable scaffolding effect: a Learner trained longer with a Guide should transfer better to a no-Guide setting, since the messages have already faded in causal importance.
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. The paper introduces a two-agent system in the BabyAI gridworld: a continuously-trained Learner and a Guide that emits fixed-length discrete messages (two tokens, three symbols) after being trained by imitation learning. The authors compare a baseline Learner with a Learner assisted by a pretrained Guide across six levels, report curriculum-transfer experiments, and analyze the emergent communication through message-action correlations, message-observation correlations, and the Causal Influence of Communication (CIC) metric. They also present an 'obedience' experiment in which the experimenters preset messages to guide a trained Learner, reporting 72–76% obedience in two levels. The central claims are that the Guide's messages speed up learning of new agents, transfer across related levels, and that the emergent language is interpretable enough for direct experimenter control.

Significance. If the claims hold, the paper would demonstrate that an emergent, unsupervised discrete communication protocol can provide reusable guidance for new agents in sequential decision tasks, and that such protocols can be partially decoded by post hoc analysis. The learning-speedup and transfer results are evaluated on new Learners, so they are not circular. The paper provides three-seed learning curves, a clear architecture, and a reasonable CIC analysis. However, the strongest novel claim—that the experimenters have 'mastered the agent language'—rests on an obedience experiment that lacks critical controls and variance reporting, and the speedup claim does not isolate the message channel from the extra capacity of the co-adapted Guide. These issues make the current evidence suggestive rather than conclusive for the paper's headline claims, but they are addressable with additional experiments.

major comments (3)
  1. [Section 4.1, Section 3.3] The claim that the pretrained Guide's messages 'cause a great increase in the learning speed of a new agent' is not isolated from the effect of additional model capacity and co-adaptation. In stage 3, the Guide's weights are not frozen (Section 3.3), so the guided Learner includes a second network whose parameters are jointly finetuned, whereas the baseline Learner does not. A control with a frozen Guide, or with a Guide whose messages are replaced by fixed or random tokens while the visual pathway remains unchanged, is needed to attribute the speedup to the communicative content rather than to the extra parameters or to the Guide's pretrained visual representations.
  2. [Section 5.4] The obedience experiment does not include a control condition in which the message channel is ablated or randomized (e.g., uniform random messages, a fixed message, or no message). Because the Learner receives the 7x7x3 visual observation at every step (Section 3.2), the 72–76% obedience could partly reflect the Learner's own policy acting on the visual input, with the preset messages merely coinciding with actions the Learner would take anyway. Additionally, the epoch is selected post hoc as the one with maximum CIC, and the obedience percentages are reported without variance across seeds or epochs. Reporting control-message obedience and error bars is necessary to support the 'we have mastered the agent language' claim.
  3. [Section 5.3, Figure 8] The statement that 'CIC tends to decrease over time' is contradicted by the GoToObj panel in Figure 8, which shows a positive correlation between training time and CIC. The text says the trend is negative if more epochs after convergence are considered, but those data are not shown. Please either provide the extended curves for GoToObj or qualify the claim to the five levels where the trend holds; the abstract and Section 5.3 present the decrease as a general finding.
minor comments (6)
  1. [Section 5.2] The reference to 'Figure A.2' for the messages-given-observation distributions is incorrect: Figure A.2 shows distributions of actions given messages, not messages given observations. The correct reference appears to be Figure 7 in the main text and Figure A.3 in the supplement.
  2. [Section 5.4, Figure 9] The text says the waltz was performed in GoToLocal, but the caption of Figure 9b says 'Waltz in PutNextLocal'. Please align the text and caption.
  3. [Table A.1] The caption contains a placeholder 'from ?' where a citation should appear; please fill in the reference.
  4. [Section 4.2] The statement that the difference between guided new Learners and pretrained Learners is 'not significant' is not supported by any reported statistical test; please specify the test used and report the relevant statistic.
  5. [Section 4.1] The learning-speedup claim is based on visual inspection of the curves; a quantitative measure such as area under the learning curve or epochs to a success-rate threshold would strengthen the claim.
  6. [Figure 3] The captions say 'results per run are shown by shaded lines' but it is unclear whether the solid line is the mean and the shaded region is min-max or standard deviation; please clarify in the captions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: speedup, transfer, and CIC analyses are evaluated on held-out agents; the self-referential §5.4 obedience test is a validity limitation, not a construction-level reduction.

full rationale

The paper's central empirical claims are not circular. In §4.1–4.3, the learning-speedup and curriculum-transfer results are measured by validation success rate of new Learners, with the pretrained Guide either finetuned or transferred across levels; these comparisons are genuine out-of-sample evaluations and do not reuse fitted quantities as predictions. The CIC analysis in §5.3 applies an external metric (Lowe et al. 2019) to validation frames and reports a measured decrease over training, which is an empirical observation rather than an input to the method. The only potentially self-referential element is §5.4, where the message-action codebook is computed from the same trained Learner-Guide system, and the test epoch is selected by maximum CIC. The resulting 72–76% obedience therefore demonstrates internal consistency of the learned protocol rather than externally grounded semantics. However, this is not circular by construction: the paper's own text acknowledges that 'the observational input always remains important, especially when random messages are sent,' and nothing in the architecture or loss forces the hijacked messages to be followed. The obedience values could have been low under the distribution shift; the lack of random-message and no-message controls is a confound that weakens external validity, but it does not make the derivation equivalent to its inputs. There are no load-bearing self-citations, no imported uniqueness theorems, and no fitted parameter renamed as a prediction.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or ontological entities. The Guide is an architectural role built from existing components. The central claims rest on the hand-chosen channel size (2 tokens from 3 symbols), the imitation-learning setup, the BabyAI environment, and the CIC metric.

free parameters (2)
  • message_length = 2
    The Guide's channel is restricted to exactly two tokens per step; no ablation on message length is reported, so the expressive capacity and interpretability results depend on this arbitrary choice.
  • vocabulary_size = 3
    Three tokens yielding nine messages are chosen by hand; no sweep is shown, and the emergent language's structure is conditional on this vocabulary size.
assumptions (4)
  • domain assumption Behavioral cloning from PPO expert demonstrations defines the target behavior for all agents.
    Section 3.3 describes three training stages built on an RL expert; if the expert policy is not representative of interactive teaching, the emergent guidance is only meaningful within this demonstration distribution.
  • domain assumption The BabyAI skill hierarchy is a valid model of task relatedness and transfer.
    Cross-level generalization is interpreted through shared competencies such as ROOM and DISTR in Table A.1, so the transfer conclusions inherit this assumption.
  • domain assumption One-step Causal Influence of Communication (CIC), as defined by Lowe et al. (2019), is a valid measure of message influence.
    The paper uses one-step mutual information between messages and next actions as 'positive listening'; this assumes the mutual information reflects causal effect rather than correlation.
  • standard math The straight-through Gumbel softmax provides a usable gradient estimator for the discrete bottleneck.
    This is a standard technique from Bengio et al. (2013) and Jang et al. (2017), so it is not a central source of risk.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mastering emergent language: learning to guide in simulated navigation." pith.science (2026). https://pith.science/paper/X64ZG5MZ

@misc{pith2026190805135,
  author       = {Pith},
  title        = {Pith review of: Mastering emergent language: learning to guide in simulated navigation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X64ZG5MZ}},
  note         = {Machine review of arXiv:1908.05135}
}
read the original abstract

To cooperate with humans effectively, virtual agents need to be able to understand and execute language instructions. A typical setup to achieve this is with a scripted teacher which guides a virtual agent using language instructions. However, such setup has clear limitations in scalability and, more importantly, it is not interactive. Here, we introduce an autonomous agent that uses discrete communication to interactively guide other agents to navigate and act on a simulated environment. The developed communication protocol is trainable, emergent and requires no additional supervision. The emergent language speeds up learning of new agents, it generalizes across incrementally more difficult tasks and, contrary to most other emergent languages, it is highly interpretable. We demonstrate how the emitted messages correlate with particular actions and observations, and how new agents become less dependent on this guidance as training progresses. By exploiting the correlations identified in our analysis, we manage to successfully address the agents in their own language.

Figures

Figures reproduced from arXiv: 1908.05135 by the authors.

Figure 1
Figure 1. A screenshot of BabyAI level PutNextLocal [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Schematic visualization of the model archi [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Development of validation success rate when training a single Learner and a Learner assisted by a [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Development of validation success rate when training a single Learner from scratch, a Learner assisted [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Development of validation success rate when training a single Learner, a Learner assisted by a Guide [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Barplots visualizing the conditional distribu [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Barplots visualizing the conditional distribu [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Trajectories of Learners in indicated lev [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 8
Figure 8. Figure 8: Development of CIC on 500 validation sam [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 22 canonical work pages

  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]

    Yoav Artzi and Luke Zettlemoyer. 2013. Weakly supervised learning of semantic parsers for mapping instructions to actions. Transactions of the Association for Computational Linguistics, 1:49--62

    work page 2013

  4. [4]

    Dzmitry Bahdanau, Felix Hill, Jan Leike, Edward Hughes, Pushmeet Kohli, and Edward Grefenstette. 2018. Learning to follow language instructions with adversarial reward induction. arXiv preprint arXiv:1806.01946

    arXiv 2018

  5. [5]

    Yoshua Bengio, Nicholas L \'e onard, and Aaron Courville. 2013. Estimating or propagating gradients through stochastic neurons for conditional computation. arXiv preprint arXiv:1308.3432

    arXiv 2013

  6. [6]

    Simon Brodeur, Ethan Perez, Ankesh Anand, Florian Golemo, Luca Celotti, Florian Strub, Jean Rouat, Hugo Larochelle, and Aaron Courville. 2017. Home: A household multimodal environment. arXiv preprint arXiv:1711.11017

    arXiv 2017

  7. [7]

    Angelo Cangelosi, Emmanouil Hourdakis, and Vadim Tikhanoff. 2006. Language acquisition and symbol grounding transfer with neural networks and cognitive robots. In The 2006 IEEE International Joint Conference on Neural Network Proceedings, pages 1576--1582. IEEE

    work page 2006

  8. [8]

    Devendra Singh Chaplot, Kanthashree Mysore Sathyendra, Rama Kumar Pasumarthi, Dheeraj Rajagopal, and Ruslan Salakhutdinov. 2018. Gated-attention architectures for task-oriented language grounding. In Thirty-Second AAAI Conference on Artificial Intelligence

    work page 2018

Show all 43 references
  1. [9]

    David L Chen and Raymond J Mooney. 2011. Learning to interpret natural language navigation instructions from observations. In Twenty-Fifth AAAI Conference on Artificial Intelligence

    2011

  2. [10]

    Maxime Chevalier-Boisvert, Dzmitry Bahdanau, Salem Lahlou, Lucas Willems, Chitwan Saharia, Thien Huu Nguyen, and Yoshua Bengio. 2018. Babyai: First steps towards grounded language learning with a human in the loop. arXiv preprint arXiv:1810.08272

    2018 arXiv

  3. [11]

    Kyunghyun Cho, Bart Van Merri \"e nboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using rnn encoder-decoder for statistical machine translation. In Proceedings of the 2014 Conference on Empirica...

    2014

  4. [12]

    John D Co-Reyes, Abhishek Gupta, Suvansh Sanjeev, Nick Altieri, John DeNero, Pieter Abbeel, and Sergey Levine. 2018. Guiding policies with language via meta-learning. arXiv preprint arXiv:1811.07882

    2018 arXiv

  5. [13]

    Vincent Crawford. 1998. https://EconPapers.repec.org/RePEc:eee:jetheo:v:78:y:1998:i:2:p:286-298 A survey of experiments on communication via cheap talk . Journal of Economic Theory, 78(2):286--298

    1998

  6. [14]

    Hal Daum \'e , John Langford, and Daniel Marcu. 2009. Search-based structured prediction. Machine learning, 75(3):297--325

    2009

  7. [15]

    Serhii Havrylov and Ivan Titov. 2017. http://papers.nips.cc/paper/6810-emergence-of-language-with-multi-agent-games-learning-to-communicate-with-sequences-of-symbols.pdf Emergence of language with multi-agent games: Learning to communicate with sequences of symbols . In I. Guy...

    2017

  8. [16]

    Sachithra Hemachandra, Felix Duvallet, Thomas M Howard, Nicholas Roy, Anthony Stentz, and Matthew R Walter. 2015. Learning models for following natural language directions in unknown environments. In 2015 IEEE International Conference on Robotics and Automation (ICRA), pages 5...

    2015

  9. [17]

    Karl Moritz Hermann, Felix Hill, Simon Green, Fumin Wang, Ryan Faulkner, Hubert Soyer, David Szepesvari, Wojciech Marian Czarnecki, Max Jaderberg, Denis Teplyashin, et al. 2017. Grounded language learning in a simulated 3d world. arXiv preprint arXiv:1706.06551

    2017 arXiv

  10. [18]

    Sepp Hochreiter and J \"u rgen Schmidhuber. 1997. Long short-term memory. Neural computation, 9(8):1735--1780

    1997

  11. [19]

    Eric Jang, Shixiang Gu, and Ben Poole. 2017. Categorical reparameterization with gumbel-softmax. In Proceedings of the International Conference on Learning Representations

    2017

  12. [20]

    Emilio Jorge, Mikael K geb \" a ck, and Emil Gustavsson. 2016. http://arxiv.org/abs/1611.03218 Learning to play guess who? and inventing a grounded language as a consequence . CoRR, abs/1611.03218

    2016 arXiv

  13. [21]

    Angeliki Lazaridou, Alexander Peysakhovich, and Marco Baroni. 2017. https://openreview.net/forum?id=Hk8N3Sclg Multi-agent cooperation and the emergence of (natural) language . In 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, ...

    2017

  14. [22]

    Ryan Lowe, Jakob Foerster, Y-Lan Boureau, Joelle Pineau, and Yann Dauphin. 2019. http://arxiv.org/abs/1903.05168 On the pitfalls of measuring emergent communication

    2019 arXiv

  15. [23]

    Hongyuan Mei, Mohit Bansal, and Matthew R Walter. 2016. Listen, attend, and walk: Neural mapping of navigational instructions to action sequences. In Thirtieth AAAI Conference on Artificial Intelligence

    2016

  16. [24]

    Tomas Mikolov, Armand Joulin, and Marco Baroni. 2016. https://doi.org/10.1007/978-3-319-75477-2\_2 A roadmap towards machine intelligence . In Computational Linguistics and Intelligent Text Processing - 17th International Conference, CICLing 2016, Konya, Turkey, April 3-9, 201...

    2016 doi

  17. [25]

    Igor Mordatch and Pieter Abbeel. 2017. http://arxiv.org/abs/1703.04908 Emergence of grounded compositional language in multi-agent populations

    2017 arXiv

  18. [26]

    Ng and Stuart Russell

    Andrew Y. Ng and Stuart Russell. 2000. Algorithms for inverse reinforcement learning. In in Proc. 17th International Conf. on Machine Learning, pages 663--670. Morgan Kaufmann

    2000

  19. [27]

    Ethan Perez, Florian Strub, Harm De Vries, Vincent Dumoulin, and Aaron Courville. 2018. Film: Visual reasoning with a general conditioning layer. In Thirty-Second AAAI Conference on Artificial Intelligence

    2018

  20. [28]

    D. A. Pomerleau . 1991. https://doi.org/10.1162/neco.1991.3.1.88 Efficient training of artificial neural networks for autonomous navigation . Neural Computation, 3(1):88--97

    1991 doi

  21. [29]

    St \'e phane Ross, Geoffrey Gordon, and Drew Bagnell. 2011. A reduction of imitation learning and structured prediction to no-regret online learning. In Proceedings of the fourteenth international conference on artificial intelligence and statistics, pages 627--635

    2011

  22. [30]

    Stéphane Ross and Drew Bagnell. 2010. http://dblp.uni-trier.de/db/journals/jmlr/jmlrp9.html#RossB10 Efficient reductions for imitation learning. In AISTATS, volume 9 of JMLR Proceedings, pages 661--668. JMLR.org

    2010

  23. [31]

    Stuart Russell. 1998. Learning agents for uncertain environments (extended abstract). In Proceedings of the Eleventh Annual Conference on Computational Learning Theory, pages 101--103. ACM Press

    1998

  24. [32]

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347

    2017 arXiv

  25. [33]

    Brian Skyrms. 2010. https://doi.org/10.1093/acprof:oso/9780199580828.001.0001 Signals: Evolution, Learning, and Information

    2010

  26. [34]

    Sainbayar Sukhbaatar, Arthur Szlam, and Rob Fergus. 2016. http://papers.nips.cc/paper/6398-learning-multiagent-communication-with-backpropagation.pdf Learning multiagent communication with backpropagation . In D. D. Lee, M. Sugiyama, U. V. Luxburg, I. Guyon, and R. Garnett, ed...

    2016

  27. [35]

    Stefanie Tellex, Thomas Kollar, Steven Dickerson, Matthew R Walter, Ashis Gopal Banerjee, Seth Teller, and Nicholas Roy. 2011. Understanding natural language commands for robotic navigation and mobile manipulation. In Twenty-Fifth AAAI Conference on Artificial Intelligence

    2011

  28. [36]

    Reggia, Juan Uriagereka, and Gerald S

    Kyle Wagner, James A. Reggia, Juan Uriagereka, and Gerald S. Wilkinson. 2003. https://doi.org/10.1177/10597123030111003 Progress in the simulation of emergent communication and language . Adaptive Behavior, 11(1):37--69

    2003 doi

  29. [37]

    Sida I Wang, Percy Liang, and Christopher D Manning. 2016. Learning language games through interaction. arXiv preprint arXiv:1606.02447

    2016 arXiv

  30. [38]

    Edward C Williams, Nakul Gopalan, Mine Rhee, and Stefanie Tellex. 2018. Learning to parse natural language to grounded reward functions with weak supervision. In 2018 IEEE International Conference on Robotics and Automation (ICRA), pages 1--7. IEEE

    2018

  31. [39]

    Terry Winograd. 1972. Understanding natural language. Cognitive psychology, 3(1):1--191

    1972

  32. [40]

    Wooldridge

    Michael J. Wooldridge. 2009. An Introduction to MultiAgent Systems (2. ed.) . Wiley

    2009

  33. [41]

    Yi Wu, Yuxin Wu, Georgia Gkioxari, and Yuandong Tian. 2018. Building generalizable agents with a realistic and rich 3d environment. arXiv preprint arXiv:1801.02209

    2018 arXiv

  34. [42]

    Haonan Yu, Haichao Zhang, and Wei Xu. 2018. Interactive grounded language acquisition and generalization in a 2d world. arXiv preprint arXiv:1802.01433

    2018 arXiv

  35. [43]

    Brian D Ziebart, Andrew L Maas, J Andrew Bagnell, and Anind K Dey. 2008. Maximum entropy inverse reinforcement learning. In Aaai, volume 8, pages 1433--1438. Chicago, IL, USA

    2008

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.