Pith. sign in

REVIEW 3 major objections 4 minor 51 references

Interactive Language Learning by Question Answering

T0 review · 3 major / 4 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 This paper argues that machine reading comprehension is missing its interactive, information-seeking component, and proposes QAit, a text-based benchmark where an agent must explore a generated world before answering questions about the…

desk verdict A useful, reproducible benchmark for interactive text-based QA, but the procedural-knowledge claim for attribute questions is under-verified because observation strings may leak attribute values. read the letter →

arxiv 1908.10909 v1 pith:3JPSISID submitted 2019-08-28 cs.CL cs.LG

Xingdi Yuan , Marc-Alexandre Cote , Jie Fu , Zhouhan Lin , Christopher Pal , Yoshua Bengio , Adam Trischler This is my paper · ORCID
classification cs.CLcs.LG
keywords QuestionAnsweringwithInteractiveTextproceduralknowledgetext-basedgamesmachinereadingcomprehensiondeepreinforcementlearninginformationseekinggeneralization
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 introduces a question-answering benchmark in which a model must explore a partially observable text-based environment—moving, taking, eating, and cutting objects—before answering questions about where objects are, whether they exist, and what attributes they have. It argues that standard machine reading benchmarks omit the information-seeking component of comprehension, because static documents place all necessary text in the reader's hands and reward shallow word matching. The central empirical claim is that current deep reinforcement learning agents solve these interactive tasks barely better than chance, especially on attribute questions, while human players answer nearly perfectly. If true, the real bottleneck in reading comprehension is not extracting declarative facts from text but learning the procedure for gathering evidence, and QAit provides a controlled setting in which that procedure can be studied.

What carries the argument

The central object is QAit (Question Answering with Interactive Text), a generated corpus of text-based games with three question types—location, existence, and attribute—and two difficulty levels, one with a fixed map and one with a randomly sampled map. The mechanism that carries the argument is the interaction loop: at each turn the agent receives an observation string and feedback from its previous command, chooses an action, modifier, and object from small lexicons, and finally issues a wait command to stop and answer. Because questions are generated with access to the underlying game state, ground-truth answers are known, and a 'sufficient information bonus' measures whether the agent has actually seen the evidence needed to answer. For location questions that means observing the queried object; for attribute questions it means executing a small set of probing commands, such as trying to cut an object to test sharpness. This bonus is the diagnostic that exposes memorization, since a model can be correct on training questions while having gathered no information at all.

What would settle it

Sample a batch of QAit games and scan the raw observation strings for attribute-denoting words (sharp, edible, drinkable, cookable, openable, portable, heat source) associated with the question's target object. If such words appear before any interaction, a static reading model could answer attribute questions above chance, and the reported human-machine gap would reflect surface leakage rather than a procedural reasoning deficit.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the interactive information-gathering step, not the reading step, is where current models fail. When the same game is flattened into a static document that contains all observations a successful trajectory would see, an off-the-shelf reading model answers 86 to 93 percent of test questions across the three types. When the model has to issue text commands to reveal those observations itself, accuracy on the same question types falls to roughly the level of random guessing, and no baseline significantly beats chance on attribute questions even when trained on unlimited procedurally generated games. Humans who play the games score near perfect on location and existence questions and 75 percent on attribute questions. The failure mode is visible in the sufficient-information bonus: agents can reach 100 percent training accuracy while their bonus stays near zero, meaning they are memorizing question-answer pairs without ever acquiring the information the question requires.

Load-bearing premise

The load-bearing premise is that the observation strings never state the attribute values being asked about—for instance, a knife is never described as sharp in the room text—so an agent truly must perform actions to discover the answer; the paper provides no analysis or ablation confirming this.

Editorial extensions

If this is right

  • Models that score well on static reading benchmarks should not be credited with comprehension in settings where evidence must be actively sought; QAit separates the extraction skill from the search-and-gather skill.
  • The near-chance attribute-question performance implies that current deep reinforcement learning agents have not learned the information-seeking subroutine that human players use, even with unlimited generated training games.
  • The strong full-information result implies the difficulty is the interaction procedure itself, not the reading or answer module, so progress on QAit requires advancing the information-gathering policy rather than the encoder.
  • The memorization success on small training sets implies that finite static corpora cannot induce procedural knowledge; on-the-fly generation is necessary, but the paper shows it is not sufficient with current algorithms.

Reading between the lines

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

  • Beyond the paper: the random made-up object names are the likely load-bearing device that forces interaction on attribute questions; a direct ablation that swaps real names into the observation text would test whether the measured difficulty is procedural or an artifact of naming.
  • Beyond the paper: the sufficient-information bonus could serve as a predictive diagnostic, since the paper's results show it diverges from training accuracy before test performance does; future agents could be selected or early-stopped on this signal.
  • Beyond the paper: because the games are generated from a grammar, the same benchmark could be extended with multi-step cooking or tool-use chains, which would make the procedural knowledge required more compositional and could be used to measure how agent capability scales with procedure length.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper introduces QAit, a text-based interactive question answering task built with TextWorld. An agent must issue text commands to navigate a partially observable environment and gather information needed to answer location, existence, and attribute questions. The authors propose a QA-DQN baseline with DQN, DDQN, and Rainbow variants, compare against a random baseline and human participants, and report that RL agents perform near chance on attribute questions across training settings while humans achieve high accuracy. The paper also includes a full-information control showing that a standard MRC model (Match-LSTM) answers all question types well when a concatenated observation document is provided, and a sufficient-information analysis for location questions.

Significance. If the central claims are supported, QAit is a valuable and reproducible benchmark for studying procedural knowledge and interactive information seeking, with a public dataset and code, on-the-fly generation of training games, and a held-out zero-shot evaluation. The paper's evidence that current deep RL agents fail to generalize on attribute questions is useful and well documented, and the full-information control is a fair diagnostic. However, the benchmark's procedural-knowledge interpretation for attribute questions rests on an unverified premise about observation-string leakage, and the human-ease claim is based on a very small sample. These issues materially affect the interpretation of the headline results, though they appear fixable with additional analysis rather than requiring a change of scope.

major comments (3)
  1. [Sec. 2.3.1, Sec. 2.4, App. D] The claim that attribute answers require interaction is not verified against leakage in the generated observation strings. The paper states that object names are replaced with made-up words for attribute questions, but it does not report whether attribute-bearing descriptions (e.g., 'open' for openable, 'diced' for cuttable/processed, 'sharp' for sharp, or 'It is edible') are excluded from TextWorld's templated observations. Table 1 itself contains 'There is an open screen door' and 'a diced potato.' If any of the nine attributes in Table 2 is directly readable from the observation string, a static reading model that has located the object could answer attribute questions without executing eat/take/cut/cook, and the reported near-random attribute accuracies would reflect exploration and RL sample complexity rather than the absence of declarative information. The heuristic conditions in Appendix D, which assume the agent must attempt cut/eat/take to determine attributes, would be unnecessary for leaked attributes. Please provide an audit of observation strings (e.g., the frequency of attribute-relevant adjectives in observations versus command feedback) or an ablation in which a non-interactive baseline reads only the concatenation of observation strings (without command feedback) and attempts to answer attribute questions.
  2. [Sec. 3.2, Table 4] The human baseline is measured on only 3 participants per condition and reported as a point estimate with no variance or significance test. With 3 participants, the random-map attribute accuracy of 0.750 is statistically indistinguishable from chance, and the perfect 1.000 scores carry wide error bars. The abstract's claim that humans solve QAit 'with relative ease' is load-bearing for the benchmark's validity and should be supported by a larger participant pool with per-participant accuracy distributions and a test against chance performance.
  3. [Sec. 4, Tables 4, 9, 10] Zero-shot test performance is reported for only a single seed per condition: the model with the random seed yielding the highest training accuracy is selected for evaluation. This selection procedure biases the reported test accuracy upward and makes the near-chance attribute results more difficult to interpret, since the reported number may be the best of three rather than a typical run. Please report mean test accuracy plus standard error (and, ideally, all seed results) for the held-out sets under each training setting.
minor comments (4)
  1. [Sec. 2.4] The statement that answers are 'one word' is slightly misleading for location questions, where the answer can be any word in an observation string, but must be a container or holder name; this could be clarified in the text.
  2. [Sec. 4.4] The full-information setup relies on a 'heuristic trajectory' that is guaranteed to observe sufficient information, but the construction of this trajectory is not described in the main text; please specify how it is generated and whether it includes only observations or also command feedback.
  3. [Table 3] The statistics in Table 3 are averages over 10,000 games, but no variance or range is reported for the random-map settings; adding interquartile ranges or standard deviations would help readers gauge the spread of game difficulty.
  4. [Sec. 6] The limitation paragraph acknowledges templated English but does not mention the potential leakage of attribute information in observations; this should be explicitly listed as a limitation and addressed by the audit suggested above.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: QAit is a held-out benchmark with independently evaluated baselines; no prediction reduces to its input by construction.

full rationale

The paper's central claim is that RL baselines fail on QAit's zero-shot held-out games while humans succeed. This claim is supported by a genuine train/test split (Section 2.5: 'we provide 500 held out games that are never seen during training'), and the reward-shaping bonuses are explicitly restricted to training (Section 3.3.1: 'both used for training the command generator. Note that these rewards are part of the design of QA-DQN, but are not used to evaluate its performance'). The task itself is defined by a generator (Côté et al., 2018) whose templates are acknowledged as a limitation (Section 6: 'QAit has the obvious limitation of using templated English'); this is a data-construction choice, not a fitted parameter masquerading as a prediction. The 'Full Information Setup' (Section 4.4) is a negative control that concatenates guaranteed-sufficient observations into a static document, showing that Match-LSTM succeeds once interaction is removed; this is a deliberate ablation, not a circular derivation. The authors' self-citations (TextWorld, episodic counting reward, QA architectures) are used as tools or prior baselines, and the load-bearing result—the accuracy gap between humans and agents—does not reduce to any of those citations. The reader-identified concern that observation strings might leak attribute values (e.g., 'a sharp knife') is an unverified internal premise about the data, which would be a correctness or dataset-quality risk, not a circularity: the paper does not define 'attribute' in terms of the model's output or fit any parameter to the zero-shot answers.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central empirical claim rests on the integrity of the TextWorld generator, the correctness of state-derived QA annotations, the non-repetition of unlimited training games, and the absence of answer leakage in the observation text. These are domain assumptions the paper asserts rather than verifies. The hand-chosen design parameters (interaction budget, training set sizes, reward coefficients) affect the difficulty and baseline results, but they are not fitted to produce the claim. No new entities are introduced.

free parameters (3)
  • Maximum interaction steps = 80
    Section 2.4: 'we use 80 steps in all experiments.' This hand-chosen cap affects task difficulty and the baselines' ability to gather information; it is not fitted, but it is a free design parameter of the benchmark.
  • Attribute reward shaping coefficients = 0.1 added to sufficient-info bonus
    Appendix A.4: extra rewards for attribute questions are added 'both with coefficient of 0.1.' Hand-tuned training signal for QA-DQN; does not enter evaluation, but affects baseline results.
  • Finite training set sizes = [1, 2, 10, 100, 500]
    Section 2.5: five finite training sets are provided. These sizes are arbitrary and strongly affect the observed overfitting behavior.
assumptions (5)
  • domain assumption The TextWorld engine generates games whose dynamics and text faithfully match the intended object-attribute rules.
    Section 2.1 asserts game dynamics arise from attributes and rules, with no formal verification or exhaustive correctness audit of generated games.
  • domain assumption Ground-truth question answers computed from the underlying game state are correct.
    Section 2.3 says questions are generated 'using the game information described above'; the paper does not validate annotations against human judgments.
  • domain assumption In the unlimited setting, training games are effectively unique, preventing rote memorization.
    Section 2.5 states more than 10^40 possible games from the fixed map distribution, but no collision analysis is provided.
  • domain assumption Observation strings do not encode attribute answers, so interaction is required.
    Sections 2.1 and 2.3.1 assume the agent must interact to discover attributes; no analysis or ablation confirms that attribute values are absent from templated observations.
  • domain assumption Human results from 21 participants (3 per condition) are representative of human ability on QAit.
    Section 3.2 reports the human baseline with no variance, protocol, or practice details; the support for 'humans solve with ease' rests on this small sample.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Interactive Language Learning by Question Answering." pith.science (2026). https://pith.science/paper/3JPSISID

@misc{pith2026190810909,
  author       = {Pith},
  title        = {Pith review of: Interactive Language Learning by Question Answering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3JPSISID}},
  note         = {Machine review of arXiv:1908.10909}
}
read the original abstract

Humans observe and interact with the world to acquire knowledge. However, most existing machine reading comprehension (MRC) tasks miss the interactive, information-seeking component of comprehension. Such tasks present models with static documents that contain all necessary information, usually concentrated in a single short substring. Thus, models can achieve strong performance through simple word- and phrase-based pattern matching. We address this problem by formulating a novel text-based question answering task: Question Answering with Interactive Text (QAit). In QAit, an agent must interact with a partially observable text-based environment to gather information required to answer questions. QAit poses questions about the existence, location, and attributes of objects found in the environment. The data is built using a text-based game generator that defines the underlying dynamics of interaction with the environment. We propose and evaluate a set of baseline models for the QAit task that includes deep reinforcement learning agents. Experiments show that the task presents a major challenge for machine reading systems, while humans solve it with relative ease.

Figures

Figures reproduced from arXiv: 1908.10909 by the authors.

Figure 1
Figure 1. Overall architecture of our baseline agent. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. shows the training curves for the neural baseline agents when trained using 10 games, 500 games and the “unlimited” games settings [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Training accuracy on the random map setup. Upper row: 10 games; middle row: 500 games; lower row: unlimited games. random. This suggests the agents memorize game￾question-answer triples when data size is small, and fail to do so in unlimited games setting. This can also be observed in Appendix E, where in at￾tribute question experiments, the training accuracy is high, and sufficient information bonus is low (even cl… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Training accuracy over episodes on location questions. Upper row: fixed map, 1/2/10/100/500/unlimited games; Lower row: random map, 2/10/100/500/unlimited games [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]
Figure 5
Figure 5. Figure 5: Training accuracy over episodes on existence questions. Upper row: fixed map, 1/2/10/100/500/unlim￾ited games; Lower row: random map, 2/10/100/500/unlimited games [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]
Figure 6
Figure 6. Figure 6: Training accuracy over episodes on attribute questions. Upper row: fixed map, 1/2/10/100/500/unlim￾ited games; Lower row: random map, 2/10/100/500/unlimited games [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 14 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]

    and Riedl, M

    Ammanabrolu, P. and Riedl, M. (2019a). Playing text-adventure games with graph-based deep reinforcement learning. In NAACL , pages 3557--3565, Minneapolis, Minnesota. Association for Computational Linguistics

    work page 2019

  4. [4]

    Transfer in Deep Reinforcement Learning using Knowledge Graphs

    Ammanabrolu, P. and Riedl, M. (2019b). Transfer in deep reinforcement learning using knowledge graphs. CoRR , abs/1908.06556

    work page Pith review arXiv 2019

  5. [5]

    Anand, A., Belilovsky, E., Kastner, K., Larochelle, H., and Courville, A. C. (2018). Blindfold baselines for embodied QA . CoRR , abs/1811.05013

    arXiv 2018

  6. [6]

    Bachman, P., Sordoni, A., and Trischler, A. (2016). Towards information-seeking agents. CoRR , abs/1612.02605

    work page Pith review arXiv 2016

  7. [7]

    Berant, J., Chou, A., Frostig, R., and Liang, P. (2013). Semantic parsing on freebase from question-answer pairs. In EMNLP , pages 1533--1544. ACL

    work page 2013

  8. [8]

    Brodeur, S., Perez, E., Anand, A., Golemo, F., Celotti, L., Strub, F., Rouat, J., Larochelle, H., and Courville, A. C. (2017). Home: a household multimodal environment. CoRR , abs/1711.11017

    arXiv 2017

Show all 51 references
  1. [9]

    Chen, D., Fisch, A., Weston, J., and Bordes, A. (2017). Reading wikipedia to answer open-domain questions. CoRR , abs/1704.00051

    2017 arXiv

  2. [10]

    H., and Bengio, Y

    Chevalier - Boisvert, M., Bahdanau, D., Lahlou, S., Willems, L., Saharia, C., Nguyen, T. H., and Bengio, Y. (2018). Babyai: First steps towards grounded language learning with a human in the loop. CoRR , abs/1810.08272

    2018 arXiv

  3. [11]

    Choi, E., He, H., Iyyer, M., Yatskar, M., Yih, W., Choi, Y., Liang, P., and Zettlemoyer, L. (2018). Quac : Question answering in context. CoRR , abs/1808.07036

    2018 arXiv

  4. [12]

    Clark, P., Cowhey, I., Etzioni, O., Khot, T., Sabharwal, A., Schoenick, C., and Tafjord, O. (2018). Think you have solved question answering? try arc, the AI2 reasoning challenge. CoRR , abs/1803.05457

    2018 arXiv

  5. [13]

    Cobbe, K., Klimov, O., Hesse, C., Kim, T., and Schulman, J. (2018). Quantifying generalization in reinforcement learning. CoRR , abs/1812.02341

    2018 arXiv

  6. [14]

    E., Adada, M., Tay, W., and Trischler, A

    C\^ot\'e, M.-A., K\'ad\'ar, A., Yuan, X., Kybartas, B., Barnes, T., Fine, E., Moore, J., Hausknecht, M., Asri, L. E., Adada, M., Tay, W., and Trischler, A. (2018). Textworld: A learning environment for text-based games. CoRR , abs/1806.11532

    2018 arXiv

  7. [15]

    Das, A., Datta, S., Gkioxari, G., Lee, S., Parikh, D., and Batra, D. (2017). Embodied question answering. CoRR , abs/1711.11543

    2017 arXiv

  8. [16]

    Das, R., Munkhdalai, T., Yuan, X., Trischler, A., and McCallum, A. (2018). Building dynamic knowledge graphs from text using machine reading comprehension. CoRR , abs/1810.05682

    2018 arXiv

  9. [17]

    Devlin, J., Chang, M., Lee, K., and Toutanova, K. (2018). BERT: pre-training of deep bidirectional transformers for language understanding. CoRR , abs/1810.04805

    2018 arXiv

  10. [18]

    U., Cirik, V., and Cho, K

    Dunn, M., Sagun, L., Higgins, M., G \" u ney, V. U., Cirik, V., and Cho, K. (2017). Searchqa: A new q & a dataset augmented with context from a search engine. CoRR , abs/1704.05179

    2017 arXiv

  11. [19]

    G., Piot, B., Menick, J., Osband, I., Graves, A., Mnih, V., Munos, R., Hassabis, D., Pietquin, O., Blundell, C., and Legg, S

    Fortunato, M., Azar, M. G., Piot, B., Menick, J., Osband, I., Graves, A., Mnih, V., Munos, R., Hassabis, D., Pietquin, O., Blundell, C., and Legg, S. (2017). Noisy networks for exploration. CoRR , abs/1706.10295

    2017 arXiv

  12. [20]

    Gordon, D., Kembhavi, A., Rastegari, M., Redmon, J., Fox, D., and Farhadi, A. (2017). IQA: visual question answering in interactive environments. CoRR , abs/1712.03316

    2017 arXiv

  13. [21]

    Hausknecht, M., Loynd, R., Yang, G., Swaminathan, A., and Williams, J. D. (2019). Nail: A general interactive fiction agent

    2019

  14. [22]

    G., and Silver, D

    Hessel, M., Modayil, J., van Hasselt, H., Schaul, T., Ostrovski, G., Dabney, W., Horgan, D., Piot, B., Azar, M. G., and Silver, D. (2017). Rainbow: Combining improvements in deep reinforcement learning. CoRR , abs/1710.02298

    2017 arXiv

  15. [23]

    and Liang, P

    Jia, R. and Liang, P. (2017). Adversarial examples for evaluating reading comprehension systems. arXiv preprint arXiv:1707.07328

    2017 arXiv

  16. [24]

    S., and Zettlemoyer, L

    Joshi, M., Choi, E., Weld, D. S., and Zettlemoyer, L. (2017). Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension. CoRR , abs/1705.03551

    2017 arXiv

  17. [25]

    R., Bontrager, P., Khalifa, A., Togelius, J., and Risi, S

    Justesen, N., Torrado, R. R., Bontrager, P., Khalifa, A., Togelius, J., and Risi, S. (2018). Procedural level generation improves generality of deep reinforcement learning. CoRR , abs/1806.10729

    2018 arXiv

  18. [26]

    P., Littman, M

    Kaelbling, L. P., Littman, M. L., and Cassandra, A. R. (1998). Planning and acting in partially observable stochastic domains. Artificial intelligence , 101(1-2):99--134

    1998

  19. [27]

    Kingma, D. P. and Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980

    2014 arXiv

  20. [28]

    Kuhlthau, C. (2004). Seeking Meaning: A Process Approach to Library and Information Services . Information management, policy, and services. Libraries Unlimited

    2004

  21. [29]

    N., Jones, L., Chang, M.-W., Dai, A., Uszkoreit, J., Le, Q., and Petrov, S

    Kwiatkowski, T., Palomaki, J., Redfield, O., Collins, M., Parikh, A., Alberti, C., Epstein, D., Polosukhin, I., Kelcey, M., Devlin, J., Lee, K., Toutanova, K. N., Jones, L., Chang, M.-W., Dai, A., Uszkoreit, J., Le, Q., and Petrov, S. (2019). Natural questions: a benchmark for...

    2019

  22. [30]

    Mikolov, T., Grave, E., Bojanowski, P., Puhrsch, C., and Joulin, A. (2018). Advances in pre-training distributed word representations. In Proceedings of the International Conference on Language Resources and Evaluation (LREC 2018)

    2018

  23. [31]

    A., Veness, J., Bellemare, M

    Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., Graves, A., Riedmiller, M., Fidjeland, A. K., Ostrovski, G., et al. (2015). Human-level control through deep reinforcement learning. Nature , 518(7540):529--533

    2015

  24. [32]

    Narasimhan, K., Kulkarni, T., and Barzilay, R. (2015). Language understanding for text-based games using deep reinforcement learning. arXiv preprint arXiv:1506.08941

    2015 arXiv

  25. [33]

    Nguyen, T., Rosenberg, M., Song, X., Gao, J., Tiwary, S., Majumder, R., and Deng, L. (2016). MS MARCO: A human generated machine reading comprehension dataset. CoRR , abs/1611.09268

    2016 arXiv

  26. [34]

    Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., and Lerer, A. (2017). Automatic differentiation in pytorch. In NIPS-W

    2017

  27. [35]

    Rajpurkar, P., Zhang, J., Lopyrev, K., and Liang, P. (2016). Squad: 100, 000+ questions for machine comprehension of text. CoRR , abs/1606.05250

    2016 arXiv

  28. [36]

    Reddy, S., Chen, D., and Manning, C. D. (2018). Coqa: A conversational question answering challenge. CoRR , abs/1808.07042

    2018 arXiv

  29. [37]

    Rychalska, B., Basaj, D., Biecek, P., and Wroblewska, A. (2018). Does it care what you asked? understanding importance of verbs in deep learning qa system

    2018

  30. [38]

    Schmidhuber, J. (2010). Formal theory of creativity, fun, and intrinsic motivation (1990&\#x2013;2010). IEEE Trans. on Auton. Ment. Dev. , 2(3):230--247

    2010

  31. [39]

    Shi, T., Karpathy, A., Fan, L., Hernandez, J., and Liang, P. (2017). World of bits: An open-domain platform for web-based agents. In Precup, D. and Teh, Y. W., editors, Proceedings of the 34th International Conference on Machine Learning , volume 70 of Proceedings of Machine L...

    2017

  32. [40]

    K., Greff, K., and Schmidhuber, J

    Srivastava, R. K., Greff, K., and Schmidhuber, J. (2015). Highway networks. CoRR , abs/1505.00387

    2015 arXiv

  33. [41]

    and Berant, J

    Talmor, A. and Berant, J. (2018). The web as a knowledge-base for answering complex questions. CoRR , abs/1803.06643

    2018 arXiv

  34. [42]

    Trischler, A., Wang, T., Yuan, X., Harris, J., Sordoni, A., Bachman, P., and Suleman, K. (2016). Newsqa: A machine comprehension dataset. CoRR , abs/1611.09830

    2016 arXiv

  35. [43]

    and Jiang, J

    Wang, S. and Jiang, J. (2016). Machine comprehension using match-lstm and answer pointer. CoRR , abs/1608.07905

    2016 arXiv

  36. [44]

    Wang, W., Yang, N., Wei, F., Chang, B., and Zhou, M. (2017). Gated self-matching networks for reading comprehension and question answering. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages 189--198. Asso...

    2017

  37. [45]

    Watkins, C. J. C. H. and Dayan, P. (1992). Q-learning. Machine Learning , 8(3):279--292

    1992

  38. [46]

    Welbl, J., Stenetorp, P., and Riedel, S. (2017). Constructing datasets for multi-hop reading comprehension across documents. CoRR , abs/1710.06481

    2017 arXiv

  39. [47]

    M., van Merriënboer, B., Joulin, A., and Mikolov, T

    Weston, J., Bordes, A., Chopra, S., Rush, A. M., van Merriënboer, B., Joulin, A., and Mikolov, T. (2015). Towards ai-complete question answering: A set of prerequisite toy tasks. CoRR

    2015

  40. [48]

    W., Salakhutdinov, R., and Manning, C

    Yang, Z., Qi, P., Zhang, S., Bengio, Y., Cohen, W. W., Salakhutdinov, R., and Manning, C. D. (2018). Hotpotqa: A dataset for diverse, explainable multi-hop question answering. CoRR , abs/1809.09600

    2018 arXiv

  41. [49]

    and May, J

    Yin, X. and May, J. (2019). Learn how to cook a new recipe in a new house: Using map familiarization, curriculum learning, and common sense to learn families of text-based adventure games. CoRR , abs/1908.04777

    2019 arXiv

  42. [50]

    W., Dohan, D., Luong, M.-T., Zhao, R., Chen, K., Norouzi, M., and Le, Q

    Yu, A. W., Dohan, D., Luong, M.-T., Zhao, R., Chen, K., Norouzi, M., and Le, Q. V. (2018). Qanet: Combining local convolution with global self-attention for reading comprehension. arXiv preprint arXiv:1804.09541

    2018 arXiv

  43. [51]

    T., Hausknecht, M

    Yuan, X., C \^ o t \' e , M., Sordoni, A., Laroche, R., des Combes, R. T., Hausknecht, M. J., and Trischler, A. (2018). Counting to explore and generalize in text-based games. CoRR , abs/1806.11525

    2018 arXiv

Pith tools

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