Papers › Transformer-based World Models Are Happy With 100k Interactions

Transformer-based World Models Are Happy With 100k Interactions

13 Mar 2023arXiv:2303.07109archive 2025-07-28

Jan Robine, Marc Höftmann, Tobias Uelwer, Stefan Harmeling

Deep neural networks have been successful in many reinforcement learning settings. However, compared to human learners they are overly data hungry. To build a sample-efficient world model, we apply a transformer to real-world episodes in an autoregressive manner: not only the compact latent states and the taken actions but also the experienced or predicted rewards are fed into the transformer, so that it can attend flexibly to all three modalities at different time steps. The transformer allows our world model to access previous states directly, instead of viewing them through a compressed recurrent state. By utilizing the Transformer-XL architecture, it is able to learn long-term dependencies while staying computationally efficient. Our transformer-based world model (TWM) generates meaningful, new experience, which is used to train a policy that outperforms previous model-free and model-based reinforcement learning algorithms on the Atari 100k benchmark.

PaperPDFCodeCode Syntology ranMethod as code

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 25 code samples harvested from 1 repository linked to this paper: 16 ran (of which 6 constructed an object rather than computing a result; 8 with no instrument failure: 1 honoured, 1 violated, 6 with no contract checked; 8 where Syntology's instrument failed) · 9 unverified. Where Syntology's instrument failed: 7 where our contract draft was wrong; 1 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: official repository: 25 samples from 1 repository, 16 ran (8 with no instrument failure; 8 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

jrobine/twm officialmentioned in paperpytorchSyntology: 25 samples, 16 ran (8 instrument-failed) · 9 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

25 samples harvested: 16 ran (of which 6 constructed an object rather than computing a result; 8 with no instrument failure: 1 honoured, 1 violated, 6 with no contract checked; 8 where Syntology's instrument failed) · 9 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

1ran · honoured contract
1ran · violated contract
7ran · our draft was wrong
1ran; the contract check could not run
6ran
9unverified

Licence: 0 of the 25 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from jrobine/twm. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. For 1 sample that ran, the graph records a signature this page does not show: for a class the fixture built without calling, the harness called the object with inputs it made up itself, so the recorded signature is the harness's probe, not the fixture's arguments. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

jrobine/twm official repository 25 samples, 16 ran (8 instrument-failed) · 9 unverified
AdamOptim twm/world_model.py ran MIT (permissive) · 006d91bc0bd061db · report fixture: returned AdamOptim · level recorded 2026-09-22
PositionalEncoding in 126 papers
census: 13 implementations from 9 papers, 9 distinct outputs across 5 buckets, one shared input per bucket
twm/world_model.py ran fingerprinted MIT (permissive) · e1c8c42def7db040 · report fixture: returned PositionalEncoding · level recorded 2026-09-22
RelativeMultiheadSelfAttention twm/world_model.py ran MIT (permissive) · ea267d5f625dc1ca · report fixture: returned RelativeMultiheadSelfAttention · level recorded 2026-09-22
TransformerXLDecoder twm/world_model.py ran MIT (permissive) · 425555f50f17e87d · report fixture: returned TransformerXLDecoder · level recorded 2026-09-22
TransformerXLDecoderLayer twm/world_model.py ran MIT (permissive) · fdc181c488b365db · report fixture: returned TransformerXLDecoderLayer · level recorded 2026-09-22
_MultilayerModule twm/world_model.py ran MIT (permissive) · 8cfa03dabb18e3af · report fixture: returned _MultilayerModule · level recorded 2026-09-22
_calculate_gain twm/world_model.py ran · honoured contract MIT (permissive) · d25a054b95990a84 · report fixture: signature nonlinearity:str/0/str, param:NoneType/0/? · returned float · level recorded 2026-09-22
_get_initializer twm/world_model.py ran · our draft was wrong MIT (permissive) · a7250d5b635dd398 · report fixture: signature name:str/0/str, nonlinearity:str/0/str, param:NoneType/0/? · returned function · level recorded 2026-09-22
combine_metrics twm/world_model.py ran · our draft was wrong MIT (permissive) · ab154d639b4e4ae3 · report fixture: signature metrics:list/1/dict, prefix:list/1/str · returned dict · level recorded 2026-09-22
get_activation in 218 papers twm/world_model.py ran · our draft was wrong MIT (permissive) · 5439c3b62dbdc1fc · report fixture: signature nonlinearity:str/0/str, param:NoneType/0/? · returned ReLU · level recorded 2026-09-22
get_norm_1d twm/world_model.py ran · our draft was wrong MIT (permissive) · 4e5362cab16fd0f9 · report fixture: signature norm:str/0/str, k:int/0/int · returned BatchNorm1d · level recorded 2026-09-22
get_norm_2d twm/world_model.py ran · our draft was wrong MIT (permissive) · 84eae076a18b163d · report fixture: signature norm:str/0/str, c:int/0/int, h:NoneType/0/?, w:NoneType/0/? · returned BatchNorm2d · level recorded 2026-09-22
init_ in 12 papers twm/world_model.py ran · our draft was wrong MIT (permissive) · fd4dd5bdc9742fc0 · report fixture: signature mod:callable/0/?, weight_initializer:str/0/str, bias_initializer:str/0/str, nonlinearity:str/0/str, param:NoneType/0/? · returned SimpleModel · level recorded 2026-09-22
same_batch_shape twm/world_model.py ran · violated contract MIT (permissive) · 7e4ce9c6f41f3336 · report fixture: signature tensors:list/1/Tensor, ndim:int/0/int · returned bool · level recorded 2026-09-22
same_batch_shape_time_offset twm/world_model.py ran; the contract check could not run (RAISES) fingerprinted MIT (permissive) · 165f8bc0355a9865 · report fixture: signature a:Tensor/3/float, b:Tensor/3/float, offset:int/0/int · returned bool · level recorded 2026-09-22
update_metrics twm/world_model.py ran · our draft was wrong MIT (permissive) · 6f69e804fca3263d · report fixture: signature metrics:dict/0/?, new_metrics:dict/0/?, prefix:str/0/str · returned dict · level recorded 2026-09-22
CNN in 23 papers twm/world_model.py unverified MIT (permissive) · 173854afd0798188 · report
DynamicsModel twm/world_model.py unverified MIT (permissive) · c61718d72bba4f72 · report
MLP in 375 papers twm/world_model.py unverified MIT (permissive) · 103a05c5c1b23236 · report
ObservationModel in 6 papers twm/world_model.py unverified MIT (permissive) · 430454696696dc84 · report
PredictionNet twm/world_model.py unverified MIT (permissive) · 34a08d167acd0d24 · report
TransposeCNN twm/world_model.py unverified MIT (permissive) · 82ffe40472f6c427 · report
WorldModel in 13 papers twm/world_model.py unverified MIT (permissive) · dd79139c2de865e1 · report
_kaiming_uniform_ twm/world_model.py unverified MIT (permissive) · 426051c1e209bfef · report
check_no_grad twm/world_model.py unverified MIT (permissive) · 862742223d0ec96a · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it.

1 assembly on this paper: 1 harvested, 0 generated.

harvested Assembly for Transformer-based World Model (TWM) (graph method: Transformer-based World Model (TWM)). 25 bricks; individually on a fixture: 16 ran (of which 6 constructed an object rather than computing a result; 8 with no instrument failure: 1 honoured, 1 violated, 6 with no contract checked; 8 where Syntology's instrument failed) · 9 unverified. The assembly has not been run end-to-end.

  1. WorldModel entry · level 0 · unverified · dd79139c2de865e1
  2. AdamOptim helper · level 1 · ran · 006d91bc0bd061db
  3. DynamicsModel model · level 0 · unverified · c61718d72bba4f72
  4. ObservationModel model · level 0 · unverified · 430454696696dc84
  5. check_no_grad helper · level 0 · unverified · 862742223d0ec96a
  6. combine_metrics helper · level 1 · ran · our draft was wrong · ab154d639b4e4ae3
  7. same_batch_shape helper · level 1 · ran · violated contract · 7e4ce9c6f41f3336
  8. same_batch_shape_time_offset helper · level 1 · ran; the contract check could not run (RAISES) · 165f8bc0355a9865
  9. PredictionNet model · level 0 · unverified · 34a08d167acd0d24
  10. CNN helper · level 0 · unverified · 173854afd0798188
  11. MLP helper · level 0 · unverified · 103a05c5c1b23236
  12. TransposeCNN helper · level 0 · unverified · 82ffe40472f6c427
  13. update_metrics helper · level 1 · ran · our draft was wrong · 6f69e804fca3263d
  14. TransformerXLDecoder model · level 1 · ran · 425555f50f17e87d
  15. TransformerXLDecoderLayer helper · level 1 · ran · fdc181c488b365db
  16. _MultilayerModule helper · level 1 · ran · 8cfa03dabb18e3af
  17. get_norm_2d helper · level 1 · ran · our draft was wrong · 84eae076a18b163d
  18. get_norm_1d helper · level 1 · ran · our draft was wrong · 4e5362cab16fd0f9
  19. PositionalEncoding helper · level 1 · ran · e1c8c42def7db040
  20. RelativeMultiheadSelfAttention model · level 1 · ran · ea267d5f625dc1ca
  21. get_activation helper · level 1 · ran · our draft was wrong · 5439c3b62dbdc1fc
  22. init_ helper · level 1 · ran · our draft was wrong · fd4dd5bdc9742fc0
  23. _get_initializer helper · level 1 · ran · our draft was wrong · a7250d5b635dd398
  24. _calculate_gain helper · level 1 · ran · honoured contract · d25a054b95990a84
  25. _kaiming_uniform_ helper · level 0 · unverified · 426051c1e209bfef

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2303.07109") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Model-based Reinforcement LearningReinforcement LearningReinforcement Learning (RL)reinforcement-learning

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

AdamAdaptive Input RepresentationsAdaptive SoftmaxAttentionCosine AnnealingDense ConnectionsDropoutLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionReLUResidual ConnectionSoftmaxTransformer-XLVariational Dropout

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections