Learn How to Cook a New Recipe in a New House: Using Map Familiarization, Curriculum Learning, and Bandit Feedback to Learn Families of Text-Based Adventure Games

Jonathan May; Xusen Yin

REVIEW 3 major objections 5 minor 2 cited by

Learn How to Cook a New Recipe in a New House: Using Map Familiarization, Curriculum Learning, and Bandit Feedback to Learn Families of Text-Based Adventure Games

T0 review · 3 major / 5 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 shows that a deep Q-learning agent can play never-before-seen text-based cooking games, reaching 72% of possible score on one test set, by combining three human-like training strategies: curriculum learning, map…

desk verdict A plausible empirical study on zero-shot text-game generalization; the curriculum result is confounded by training compute and LinUCB needs a validity check, but the paper deserves peer review. read the letter →

arxiv 1908.04777 v3 pith:F2IYR6WU submitted 2019-08-13 cs.CL cs.AI

Xusen Yin , Jonathan May This is my paper · ORCID

classification cs.CLcs.AI

keywords text-basedgamesdeepreinforcementlearningcurriculumknowledgegraphLinUCBzero-shotgeneralizationTextWorldDRRN

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 asks whether an agent trained on a family of text-based adventure games can generalize to new, unseen instances of the same theme, rather than overfitting to a single game. It claims that three simple, human-inspired changes to a deep Q-learning recipe make this possible: training on easy games before hard ones, converting location-specific navigation knowledge into general room-type knowledge, and using a contextual bandit (LinUCB) to explore uncertain actions during evaluation. On the cooking-game test sets, the best configuration scores 72% and 68%, compared with 14% for random actions and 50–54% for training on all game types at once. The result matters because it suggests a path toward agents that can handle novel environments in open-world text tasks, closer to how people transfer cooking or navigation skills to a new kitchen.

What carries the argument

The central object is the DRRN agent, which scores each action by the bilinear form $f(s,a) = h_s^\top W h_a$, where $h_s$ is the CNN-encoded trajectory and $h_a$ is the LSTM-encoded action. Three mechanisms carry the argument: (1) a hand-designed curriculum over six game tiers that starts with single-room, one-ingredient games and progresses to twelve-room games with up to three ingredients; (2) a knowledge graph that supplies the destination room type for each navigation action, turning the instance-specific 'go east' into the universal 'go east to kitchen'; and (3) LinUCB applied at evaluation time, which solves a ridge-regression problem per action to estimate the covariance $V_a$ of seen state-action pairs and then selects the action maximizing $Q_{s,a} + c\sqrt{h_s^\top V_a^{-1} h_s}$. This bandit mechanism is what injects directed exploration without destructive random actions.

What would settle it

A direct test is to take the trained DRRN agent, collect the encoded states and Q-values it produces during evaluation, and fit the linear model $Q = h_s^\top \theta_a + \eta$ per action; if the residuals show systematic curvature or correlation with the state features beyond noise, the linearity assumption is violated. A behavioral falsification would be to replace LinUCB's confidence-bound exploration with pure random exploration over the same number of extra steps (e.g., 50 steps per episode), holding everything else fixed; if random exploration matches LinUCB's scores, the bandit's uncertainty-directed selection is not the cause of the improvement.

Watch

Extended reading notes

Core claim

The central claim is that deep Q-learning, originally designed for single video games, can be extended to zero-shot generalization across a family of text adventure games if the training and inference procedures are structured to separate universal knowledge from instance-specific knowledge. Concretely, the authors train a Deep Reinforcement Relevance Network (DRRN) on cooking games generated by the TextWorld engine, organizing games into six difficulty tiers and training sequentially (curriculum learning). They convert instance-level navigation actions like 'go east' into either a random direction ('go-random') or a room-destination-augmented action ('go room east to kitchen') using a knowledge graph built by random walking during early steps of each episode. At inference, instead of greedily picking the highest Q-value, they model each action's Q-value as a linear function of the encoded state and use LinUCB confidence bounds to select actions that explore rarely seen state-action pairs. The reported result is that the full recipe—curriculum, go-room augmentation, and LinUCB—achieves the highest aggregate scores on two held-out test sets of 222 unseen games each, with the gains most pronounced in the hardest twelve-room tier.

Load-bearing premise

The LinUCB evaluation improvement rests on the assumption that the true Q-value for each action is linearly related to the encoded state plus sub-Gaussian noise, even though the actual Q-values come from a nonlinear neural network; if this linearity fails, the confidence bounds are miscalibrated and the reported exploration gains could be accidental.

Editorial extensions

If this is right

If the result holds, training on a curriculum of carefully ordered game tiers is a practical alternative to mixing all training data, improving zero-shot performance on the same family of tasks.
Representing actions by their semantic effect (destination room type) rather than their surface form ('go east') transfers better to new environments, suggesting that universal knowledge can be encoded in action representations.
Applying LinUCB-style confidence bounds at inference time, rather than during training, improves exploration in zero-shot settings and avoids the destructive random moves that damage game progress.
The method establishes a strong baseline on the TextWorld cooking competition's zero-shot evaluation, at 72% of achievable score on unseen cooking games.

Reading between the lines

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

The go-room augmentation suggests a broader design principle: for any task with a spatial or structural component, making the action representation include the predicted next-state category may improve generalization, and this could be tested in navigation-oriented dialogue agents beyond text games.
The LinUCB gains, if they replicate, would indicate that the main bottleneck in zero-shot text-game play is not policy quality but confidence calibration over rarely seen states; a well-calibrated uncertainty estimate could be as valuable as more training data.
The reported recency effect—where the agent performs best on the tier most recently trained—implies a continual-learning failure mode; a natural extension would be to replay old-tier games during later training to preserve skills, which the authors only partially do with a final fine-tuning stage.
The assumption that the action set is fully known at inference limits the method to games with predefined admissible actions; future work could relax this by generating candidate actions from a language model, but the paper leaves this unexplored.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Short version: this is a legitimate empirical study on zero-shot generalization for families of TextWorld cooking games. The simple strategies — curriculum learning, rewriting go actions to room destinations, and LinUCB-style exploration at inference — do seem to improve task completion over reasonable baselines. But the headline curriculum result is confounded by unequal training compute, and the LinUCB improvement rests on an unvalidated linearity assumption. Worth serious review, but the authors need to run a matched-compute control and release code.

What's new: most prior text-adventure work optimizes a single game; this one evaluates on never-before-seen games of a theme. That's the right setup. The tier breakdowns and the recency analysis (Table 5) are thoughtful. I believe the go-room action rewriting is a genuinely useful trick: converting instance-level direction info into universal room-transition info helps in larger maps. The zero-shot scores are well above random (72% vs 14%) and mixed-only training (50-54%).

The soft spots are real. The stress-test note is correct: Section 5.2 says curriculum decays epsilon over up to 2M steps per stage across six tiers plus fine-tuning, while 'mixed' decays over 10M steps. So the comparison does not match total optimizer steps or exploration schedule. The observed 64% vs 50-54% could partly be training compute, not curriculum ordering. The paper needs a matched-compute curriculum baseline. Without that, the abstract's first contribution is unsupported.

LinUCB is the other place to push. The linearity assumption in Section 3.3 is stated but never checked. Q-values come from a nonlinear DRRN; the confidence bounds may be miscalibrated. The coefficient 0.2 is hand-chosen, and there's no sensitivity analysis. Table 6 shows some tier-level CIs overlap, and Table 2 lacks error bars. I'm not claiming the LinUCB result is false; it's just not established.

One more: the conclusion says bandit feedback helps 'both training and evaluation phases,' but LinUCB is only used at evaluation. That's a wording bug, not a fatal one.

This paper is for researchers working on generalization in RL for text games. It deserves a serious referee; the result is plausible and the benchmark setup is useful. Recommend acceptance after revision, with matched-compute control, significance/effect sizes, and a LinUCB robustness check.

Referee Report

3 major / 5 minor

Summary. The paper studies zero-shot generalization in families of text-based adventure games generated by the TextWorld framework. The authors propose three techniques: curriculum learning over six difficulty tiers, conversion of instance-specific navigation knowledge into universal knowledge via random-go and go-room strategies, and LinUCB-based exploration at inference time. Experiments on 4,440 cooking games report improved task completion rates on two held-out test sets: curriculum learning over mixed training (64% vs 50–54% overall), go-room over go-cardinal, and LinUCB over greedy and sampling policies (72% vs 69% on Test 1). The central claim is that these simple human-inspired strategies improve the agent's ability to play never-before-seen games of the same theme.

Significance. If the results hold, the paper is a useful empirical contribution to the under-studied problem of zero-shot generalization in text-based games, and it demonstrates that simple, well-chosen inductive biases can beat both flat training and naive stochastic evaluation. The authors use a public benchmark, hold out games by type, select models on a dev set, and evaluate on genuinely unseen games, which is a strength. The paper also provides several ablations to isolate the effect of each proposed component. However, the two most load-bearing comparisons—curriculum versus mixed training, and LinUCB versus sampling—have problems that prevent the conclusions from being fully established: the curriculum comparison is confounded with training budget and epsilon schedule, and the LinUCB method rests on an unvalidated linearity assumption that may not hold for a nonlinear deep Q-network. The statistical reporting is also incomplete, with most tables lacking error bars or significance tests.

major comments (3)

[§5.2, Tables 2–3] The headline comparison between 'curric' and 'mixed' is confounded with training budget and exploration schedule. §5.2 states that each curriculum stage decays epsilon 'across a maximum of two million steps' over six stages plus a final fine-tuning stage, while 'mix go-room' 'instead decay[s] over 10 million steps.' The curriculum condition therefore receives potentially up to 14M optimizer updates with epsilon reset at every stage, whereas the mixed condition receives 10M updates with a single decay. Because both total update count and epsilon schedule strongly influence deep Q-learning, the 14-point gap in Table 3 cannot be attributed solely to the ordering of training data. Please add a matched-compute control with identical total steps and epsilon schedule, or provide explicit evidence that the compute difference does not drive the result.
[§3.3, Table 6] The LinUCB evaluation procedure rests on the assumption that during evaluation Q_{s,a} = h_s^T θ_a + η_a with η_a R-sub-Gaussian, but h_s is produced by a CNN and the Q-values come from the nonlinear DRRN. The paper provides no validation of this linearity, so the confidence bounds used for action selection may be miscalibrated. Section 5.4 further introduces a normalization coefficient of 0.2 with no sensitivity analysis, leaving open the possibility that the reported LinUCB gains depend on this arbitrary constant. Since the improvement from LinUCB over sampling (72 vs 69 on Test 1) is the basis of the third contribution, please include fit diagnostics on held-out (h_s, Q) pairs, or explicitly reframe LinUCB as an unvalidated heuristic and investigate sensitivity to the 0.2 coefficient.
[Tables 2–6] The main numerical claims lack statistical support. Tables 2–4 report single point estimates with no variance across training runs or evaluation episodes, and the confidence intervals in Table 6 are based on only 10 episodes per game, with per-tier differences that overlap (e.g., Test 1, tier 5: sampling 70±2 vs UCB 69±0). Without error bars or significance tests on the aggregate comparisons, the claimed superiority of curriculum over mixed training and of LinUCB over sampling is not established. Please report variance across at least three training seeds and test the significance of the overall differences.

minor comments (5)

[§5.2, Table 3] The table caption says 'Rows 1-6 show the breakdown of total scores and steps,' but the table contains only scores; please remove the reference to steps or add the step counts.
[§7] The conclusion states that bandit feedback is incorporated into 'both training and evaluation phases,' but Section 3.3 applies LinUCB only at evaluation time; the conclusion should be corrected to avoid overstating the method.
[§3.3] The definition of the confidence-bound coefficient c_t is left as 'a normalization value related to V_t^a' with references to prior work; for reproducibility, please provide the explicit formula or a precise citation to the equation used.
[§3.2] The terms 'universal knowledge' and 'instance knowledge' are used implicitly before being defined; please provide explicit definitions at first use.
[Table 4] The row labels 'Test-1' and 'Test-2' are written with a hyphen, unlike the rest of the paper, which uses 'Test 1' and 'Test 2'; please make the notation consistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: all central claims are empirical comparisons against baselines, and no equation reduces to its own inputs.

full rationale

The paper's central contributions — curriculum learning, go-room map familiarization, and LinUCB-based evaluation — are each supported by direct experimental comparisons on held-out TextWorld games, not by derivation from the assumptions that define them. The curriculum-versus-mixed comparison is an empirical ablation; although Section 5.2 shows the two conditions use different total training steps and epsilon schedules, that is a potential experimental confound rather than a circular reduction. The LinUCB method in Section 3.3 assumes a linear model of Q-values and fits ridge-regression coefficients, but the action-selection rule adds the resulting confidence bound to the DRRN Q-values as an exploration bonus, and the reported scores are measured from actual game outcomes. Thus the improvement is not forced by construction, even though the linearity assumption itself is unvalidated. Self-citations to Yin and May (2019) supply the CNN encoder architecture and motivate exploration, but no load-bearing claim reduces to that citation: the new strategies are tested against random and mixed baselines, and the cited work is not used to define the outcome metric. No fitted parameter is renamed as a prediction, and no equation in the paper equals another by definition. Therefore no specific circular step can be quoted.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central empirical claims rest on standard RL hyperparameters and two explicit domain assumptions: known action sets at inference and a linear model for Q-values during LinUCB evaluation. No new physical or mathematical entities are introduced.

free parameters (6)

LinUCB bound coefficient = 0.2
Chosen to scale the normalized confidence bounds at each step; no sensitivity analysis is reported (Section 5.4).
Gumbel-max temperature = 0.01
Used for the sampling-from-policy baseline; chosen to sharpen the action distribution (Section 5.4).
Step penalty = -0.1
Reward shaping constant subtracted per step during training (Section 4.2).
Loss penalty = -1
Instant reward added when an episode ends early due to a failed action (Section 4.2).
Epsilon decay schedule = 1 to 1e-4 over 2M or 10M steps
Exploration schedule for curriculum stages versus mixed training; the two compared conditions may not have matched total training steps (Section 5.2).
Ridge regression lambda = not stated
Appears in the LinUCB objective but its value is never reported, making the bound computation incompletely specified (Section 3.3).

assumptions (4)

domain assumption The Q-value for each action is linear in the encoded state plus R-sub-Gaussian noise during evaluation.
Used to justify LinUCB confidence bounds (Section 3.3); not verified against the nonlinear DRRN Q-values.
domain assumption The action set is fully known at inference time.
Explicit simplification in Section 2.1; open text generation is out of scope.
domain assumption Game tiers reflect a valid difficulty ordering for curriculum learning.
Tiers 1 to 6 are defined by ingredient count and room count (Table 1); the paper does not independently validate this ordering.
standard math Standard linear bandit concentration inequalities apply to this setting.
Linear bandit theory from Abbasi-yadkori et al. is invoked to derive the upper confidence bound (Section 3.3).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Learn How to Cook a New Recipe in a New House: Using Map Familiarization, Curriculum Learning, and Bandit Feedback to Learn Families of Text-Based Adventure Games." pith.science (2026). https://pith.science/paper/F2IYR6WU

@misc{pith2026190804777,
  author       = {Pith},
  title        = {Pith review of: Learn How to Cook a New Recipe in a New House: Using Map Familiarization, Curriculum Learning, and Bandit Feedback to Learn Families of Text-Based Adventure Games},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F2IYR6WU}},
  note         = {Machine review of arXiv:1908.04777}
}

read the original abstract

We consider the task of learning to play families of text-based computer adventure games, i.e., fully textual environments with a common theme (e.g. cooking) and goal (e.g. prepare a meal from a recipe) but with different specifics; new instances of such games are relatively straightforward for humans to master after a brief exposure to the genre but have been curiously difficult for computer agents to learn. We find that the deep Q-learning strategies that have been successfully leveraged for superhuman performance in single-instance action video games can be applied to learn families of text video games when adopting simple strategies that correlate with human-like learning behavior. Specifically, we build agents that learn to tackle simple scenarios before more complex ones using curriculum learning, that familiarize themselves in an unfamiliar environment by navigating before acting, and that explore uncertain environments more thoroughly using contextual multi-armed bandit decision policies. We demonstrate improved task completion rates over reasonable baselines when evaluating on never-before-seen games of that theme.

Figures

Figures reproduced from arXiv: 1908.04777 by the authors.

**Figure 2.** The training process of ‘curric’ with go-room strategy bro [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Interactive Language Learning by Question Answering
cs.CL 2019-08 conditional novelty 6.0 of 10

QAit turns question answering into an interactive text-game task, and the paper's baselines show current agents cannot generalize beyond memorized games, while humans can.
Multi-Agent Language Models: Advancing Cooperation, Coordination, and Adaptation
cs.CL 2025-06 conditional novelty 3.0 of 10

A thesis proposal repurposing two prior papers on LM agents for text games, framed as a path to theory-of-mind AI, with no new theory-of-mind evidence.

Reference graph

Works this paper leans on

21 extracted references · 17 canonical work pages · cited by 2 Pith papers

[1]

Improved algorithms for lin- ear stochastic bandits

[Abbasi-yadkori et al., 2011] Yasin Abbasi-yadkori, D ´avid P´al, and Csaba Szepesv ´ari. Improved algorithms for lin- ear stochastic bandits. In J. Shawe-Taylor, R. S. Zemel, P. L. Bartlett, F. Pereira, and K. Q. Weinberger, editors, Advances in Neural Information Processing Systems 24 , pages 2312–2320. Curran Associates, Inc.,

work page 2011
[5]

Using conﬁdence bounds for exploitation-exploration trade-offs

[Auer, 2003] Peter Auer. Using conﬁdence bounds for exploitation-exploration trade-offs. J. Mach. Learn. Res., 3:397–422, March

work page 2003
[6]

Curriculum learning

[Bengio et al., 2009] Yoshua Bengio, J ´erˆome Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th Annual International Confer- ence on Machine Learning, ICML ’09, pages 41–48, New York, NY , USA,

work page 2009
[8]

What can you do with a rock? affordance extraction via word embeddings

[Fulda et al., 2017] Nancy Fulda, Daniel Ricks, Ben Mur- doch, and David Wingate. What can you do with a rock? affordance extraction via word embeddings. In Carles Sierra, editor, IJCAI, pages 1039–1045. ijcai.org,

work page 2017
[13]

Text-based ad- ventures of the golovin AI agent

[Kostka et al., 2017] Bartosz Kostka, Jaroslaw Kwiecieli, Jakub Kowalski, and Pawel Rychlikowski. Text-based ad- ventures of the golovin AI agent. In CIG, pages 181–188. IEEE,

work page 2017
[14]

Special feature zork: A computerized fantasy simulation game

[Lebling et al., 1979] Lebling, Blank, and Anderson. Special feature zork: A computerized fantasy simulation game. Computer, 12(4):51–59, April

work page 1979
[16]

[Mnih et al., 2015] V olodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A

Association for Computational Linguis- tics. [Mnih et al., 2015] V olodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A. Rusu, Joel Veness, Marc G. Belle- mare, Alex Graves, Martin Riedmiller, Andreas K. Fid- jeland, Georg Ostrovski, Stig Petersen, Charles Beattie, Amir Sadik, Ioannis Antonoglou, Helen King, Dharshan Kumaran, Daan Wierstra, Shane Legg...

work page 2015
[17]

Language understanding for text- based games using deep reinforcement learning

[Narasimhan et al., 2015] Karthik Narasimhan, Tejas Kulka- rni, and Regina Barzilay. Language understanding for text- based games using deep reinforcement learning. In Pro- ceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 1–11. Association for Computational Linguistics,

work page 2015

Show all 21 references

[18]

Gumbel-max trick and weighted reservoir sampling,

[Vieira, 2014] Tim Vieira. Gumbel-max trick and weighted reservoir sampling,

2014
[21]

Mankowitz, and Shie Mannor

[Zahavy et al., 2018] Tom Zahavy, Matan Haroush, Nadav Merlis, Daniel J. Mankowitz, and Shie Mannor. Learn what not to learn: Action elimination with deep reinforce- ment learning. In Advances in Neural Information Pro- cessing Systems 31: Annual Conference on Neural In- forma...

2018
[1979]

Deep reinforce- ment learning for dialogue generation

[Li et al., 2016] Jiwei Li, Will Monroe, Alan Ritter, Dan Ju- rafsky, Michel Galley, and Jianfeng Gao. Deep reinforce- ment learning for dialogue generation. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1192–1202, Austin, Tex...

2016
[1997]

Kingma and Jimmy Ba

[Kingma and Ba, 2015] Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In 3rd Inter- national Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings,

2015
[2003]

[Ammanabrolu and Riedl, 2018] Prithviraj Ammanabrolu and Mark O. Riedl. Playing text-adventure games with graph-based deep reinforcement learning. CoRR, abs/1812.01628,

2018 arXiv
[2009]

[Cˆot´e et al., 2018] Marc-Alexandre C ˆot´e, ´Akos K ´ad´ar, Xingdi Yuan, Ben Kybartas, Tavian Barnes, Emery Fine, James Moore, Matthew J

ACM. [Cˆot´e et al., 2018] Marc-Alexandre C ˆot´e, ´Akos K ´ad´ar, Xingdi Yuan, Ben Kybartas, Tavian Barnes, Emery Fine, James Moore, Matthew J. Hausknecht, Layla El Asri, Mahmoud Adada, Wendy Tay, and Adam Trischler. Textworld: A learning environment for text-based games. CoR...

2018 arXiv
[2011]

Biermann, and Philip M

[Abe et al., 2003] Naoki Abe, Alan W. Biermann, and Philip M. Long. Reinforcement learning with immediate rewards and linear hypotheses.Algorithmica, 37:263–293,

2003
[2014]

Com- prehensible context-driven text game playing

[Yin and May, 2019] Xusen Yin and Jonathan May. Com- prehensible context-driven text game playing. CoRR, abs/1905.02265,

2019 arXiv
[2015]

Long short-term memory

[Hochreiter and Schmidhuber, 1997] Sepp Hochreiter and J¨urgen Schmidhuber. Long short-term memory. Neural computation, 9:1735–80, 12

1997
[2016]

Distilling the knowledge in a neural net- work

[Hinton et al., 2015] Geoffrey Hinton, Oriol Vinyals, and Jeffrey Dean. Distilling the knowledge in a neural net- work. In NIPS Deep Learning and Representation Learn- ing Workshop,

2015
[2017]

Deep reinforcement learning with a natural language ac- tion space

[He et al., 2016] Ji He, Jianshu Chen, Xiaodong He, Jian- feng Gao, Lihong Li, Li Deng, and Mari Ostendorf. Deep reinforcement learning with a natural language ac- tion space. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1:...

2016
[2018]

P, Sarath Chandar, and Balaraman Ravindran

[Ansari et al., 2018] Ghulam Ahmed Ansari, Sagar J. P, Sarath Chandar, and Balaraman Ravindran. Language expansion in text-based games. CoRR, abs/1805.07274,

2018 arXiv
[2019]

Hausknecht, and Adam Trischler

[Yuan et al., 2018] Xingdi Yuan, Marc-Alexandre C ˆot´e, Alessandro Sordoni, Romain Laroche, Remi Tachet des Combes, Matthew J. Hausknecht, and Adam Trischler. 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.

[1] [1]

Improved algorithms for lin- ear stochastic bandits

[Abbasi-yadkori et al., 2011] Yasin Abbasi-yadkori, D ´avid P´al, and Csaba Szepesv ´ari. Improved algorithms for lin- ear stochastic bandits. In J. Shawe-Taylor, R. S. Zemel, P. L. Bartlett, F. Pereira, and K. Q. Weinberger, editors, Advances in Neural Information Processing Systems 24 , pages 2312–2320. Curran Associates, Inc.,

work page 2011

[2] [5]

Using conﬁdence bounds for exploitation-exploration trade-offs

[Auer, 2003] Peter Auer. Using conﬁdence bounds for exploitation-exploration trade-offs. J. Mach. Learn. Res., 3:397–422, March

work page 2003

[3] [6]

Curriculum learning

[Bengio et al., 2009] Yoshua Bengio, J ´erˆome Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th Annual International Confer- ence on Machine Learning, ICML ’09, pages 41–48, New York, NY , USA,

work page 2009

[4] [8]

What can you do with a rock? affordance extraction via word embeddings

[Fulda et al., 2017] Nancy Fulda, Daniel Ricks, Ben Mur- doch, and David Wingate. What can you do with a rock? affordance extraction via word embeddings. In Carles Sierra, editor, IJCAI, pages 1039–1045. ijcai.org,

work page 2017

[5] [13]

Text-based ad- ventures of the golovin AI agent

[Kostka et al., 2017] Bartosz Kostka, Jaroslaw Kwiecieli, Jakub Kowalski, and Pawel Rychlikowski. Text-based ad- ventures of the golovin AI agent. In CIG, pages 181–188. IEEE,

work page 2017

[6] [14]

Special feature zork: A computerized fantasy simulation game

[Lebling et al., 1979] Lebling, Blank, and Anderson. Special feature zork: A computerized fantasy simulation game. Computer, 12(4):51–59, April

work page 1979

[7] [16]

[Mnih et al., 2015] V olodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A

Association for Computational Linguis- tics. [Mnih et al., 2015] V olodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A. Rusu, Joel Veness, Marc G. Belle- mare, Alex Graves, Martin Riedmiller, Andreas K. Fid- jeland, Georg Ostrovski, Stig Petersen, Charles Beattie, Amir Sadik, Ioannis Antonoglou, Helen King, Dharshan Kumaran, Daan Wierstra, Shane Legg...

work page 2015

[8] [17]

Language understanding for text- based games using deep reinforcement learning

[Narasimhan et al., 2015] Karthik Narasimhan, Tejas Kulka- rni, and Regina Barzilay. Language understanding for text- based games using deep reinforcement learning. In Pro- ceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 1–11. Association for Computational Linguistics,

work page 2015